Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:00
Here we are in episode 105 of
0:02
the self-hosted podcast and it is
0:05
feeling officially like
0:06
the end of summer. My
0:09
two youngest have gone back to school. It's the first
0:11
day of school today. You only say that
0:13
because you live in the frozen north. It was 97 degrees
0:17
down in Raleigh today. Okay. Yeah,
0:19
actually, you know what's funny? Last night on a whim
0:21
I pulled up Raleigh because I have like, I have like everywhere
0:24
the hosts are at stored in Carrot. And
0:26
so I was on Carrot Weather. I'm like, oh, let's check in on everybody.
0:29
Yeah, it's like, oh, Alex.
0:32
Yeah, it's 68 degrees and blue skies
0:34
here.
0:34
I'll take it. But it makes me a little sad when
0:37
summer comes to an end. I always picture summer
0:39
is like, I'm going to get all these projects done. Yeah.
0:43
Do you think that stems from summers
0:45
at school as a kid feeling like they stretch
0:48
in front of you forever,
0:50
except as an adult
0:52
life is in the way. You
0:54
get that. And also for me, it's like,
0:56
well, I'm going to have more daylight. So
0:59
if the sun's going to be up till eight or nine o'clock,
1:02
I should be able to get like twice as much stuff
1:04
done. That should happen. Right.
1:07
So, all right, I'll
1:07
plan like my big thing. And I got close.
1:10
Was I, I really started for
1:12
a moment with Brent's help to wrap my head around
1:15
all the different ways I could improve
1:17
the RV and the studio with ESP
1:19
devices, modules, and just
1:21
little things that I could get like the sonar module
1:24
or relay, you know, controls and temperature
1:27
sensors and moisture sensors and
1:29
all these little things I wanted to do for LEDs,
1:31
where I wanted to add a few more LEDs controlled by an ESP.
1:34
And
1:35
I got a lot of the kits together. I
1:37
got various parts. I got some LED light
1:39
ropes. I got some relays
1:41
and then just sort of lost momentum on all of it.
1:44
And I thought by the end of summer I'd have
1:46
I'd have like these lights installed and I'd have
1:48
some stuff implemented. And nope,
1:51
I didn't get to it. Not at all. So
1:53
I feel a little bad about that. I think the trouble with
1:55
some of those sort of more
1:58
DIY projects, like the. ESP based
2:00
stuff is a lot
2:03
of the times you don't know every single
2:06
little piece of the jigsaw that you
2:08
need
2:08
until you start getting into the project. And
2:11
then inevitably you do the calculus of
2:13
well it's $15 on
2:15
Amazon for this thing or it's $4
2:18
on Ali Express for the exact same thing.
2:20
Do I just wait three months for it to arrive on the
2:23
slow boat or do I pay
2:25
the Amazon tax.
2:26
And that can be the biggest impediment for me
2:29
to finishing some of these more bitty projects
2:31
from the ESP stuff. So
2:33
I took the route I recently ordered something called an ESP
2:36
clicker in my bonus room above my drum
2:38
set I have a skylight with an
2:41
automated VLux blind motorized solar
2:43
power blind in it. And it has a remote
2:45
control and RF remote control I think I've mentioned
2:47
it on the show before which is paired
2:50
with that blind and I don't
2:52
really want to go about hacking the RF protocol
2:54
because the VLux blinds have some kind of encryption
2:57
key rotation nonsense in them which means
3:01
not random strangers can't control
3:03
my blinds which is. Yeah I guess that makes sense. Nice
3:06
but also do we care
3:08
but OK. Not for that particular
3:10
skylight probably not and it. Yeah
3:13
this is interesting because
3:15
this is probably something that everybody
3:17
has in their home with some device like this that has a
3:19
remote or whatnot. Exactly. I'm listening. This
3:21
is your right up my alley right now.
3:23
So this guy I can't tell quite where his accents
3:25
from it sounds Eastern European
3:28
sort of Russian ish I'm not entirely sure
3:30
where this chap's from but he runs a website
3:32
that priceless toolkit which is an IoT
3:35
shop and he sells preassembled
3:37
circuit boards for all sorts of ESP
3:40
related nonsense. And this
3:42
ESP clicker has three microscopic
3:45
relays on it which can simulate up
3:48
to three different button presses
3:50
on different physical devices. So
3:53
the use case he shows in the link to YouTube video on
3:55
the product page which W a link to in the show
3:57
notes by the way is that he
3:59
has.
3:59
his coffee maker, which is not Wi-Fi
4:02
enabled. The only way of interfacing
4:04
with this physical device
4:06
is to stand in front of it with your meat sausage
4:09
and just push the button on the front
4:11
of the machine. Well, except
4:13
of course, all that's doing in reality is
4:16
bridging a contact. So all the relays
4:18
doing inside the ESP clicker is the
4:20
same thing. And obviously because it integrates
4:22
with ESPHome,
4:24
Home Assistant integration is a
4:26
mere click away. And so
4:28
I imagine you kind of have to pick
4:31
and choose the device you're wiring to. Like you'd
4:33
have to be willing to open it up and wire some contacts
4:35
on that side.
4:36
Yes, absolutely. If you're not into soldering
4:38
or into hacking potentially
4:41
very expensive devices to pieces to
4:43
integrate this thing inside of, stay away.
4:45
But for me with my V-Lux
4:48
remote, it has some surface
4:50
mount buttons on it. It's got the four, each
4:52
button has four legs on it.
4:54
So I get my multi-meter out
4:56
and I turn it into continuity mode. And
5:00
the pins of my multi-meter
5:03
are tiny. The probes,
5:05
the tips of the probes are tiny. But the pads
5:08
on these service mount component buttons are
5:10
even tinier. I can't even really
5:13
think of a actual item that
5:15
we would hold in our hands that is that small.
5:18
They are maybe the head of a sewing needle
5:20
small.
5:21
Like I'm used to soldering small stuff with racing
5:24
drones but this is like another
5:26
level down. So I'm gonna have fun soldering it. But
5:29
I did manage to get it work simply by just hot gluing.
5:32
I hot glued just what I was messing about.
5:34
I hot glued the cable from the ESP clicker
5:37
onto the button. So I wouldn't have to physically hold it and
5:40
control the blinds above
5:42
my drums from Home Assistant.
5:45
So it was pretty sweet. That is nice.
5:48
I'm gonna pick one of these up. I have a fan
5:51
I would love to control. Love, love, love
5:53
to be able to remote control with Home Assistant. Cause it's
5:55
built into the roof. It's a great exhaust fan
5:58
but you don't need it running all the time. If it gets.
5:59
it's down below a certain temperature, I could just kill it with
6:02
something like this. I could just
6:03
have that automation that turns it off.
6:06
What's particularly nice about buying this from
6:08
Priceless Toolkit, we have no affiliation with
6:10
this chap or I just found a cool project
6:13
on YouTube,
6:14
is it's a pre-built project. So he
6:16
shows you how to assemble it with all the surface mount components
6:19
on the ESP board, all the relays and stuff.
6:21
But
6:22
honestly, this board is tiny. It's probably
6:24
about the size of your index finger when
6:26
it arrives.
6:28
It's too small for me to be messing around
6:31
with. I'm good at soldering small stuff, but
6:33
this is just,
6:34
it's too much. But the
6:37
fact that you and I can just go on a website
6:39
and buy a pre-made thing as a product that
6:41
arrives with ESP home
6:43
already flashed on it,
6:45
man, that is compelling. You
6:47
wish, I know it's never gonna happen, Alex, but
6:49
don't you just wish these vendors would start
6:52
maybe selling this as an option, like
6:55
an upgrade option. Hey, buy it with the ESP
6:57
home wired in and you can do what you want with
6:59
it. I would pay more for that. Well, I suppose
7:01
effectively, that's what the whole
7:03
two year thing in a roundabout
7:06
kind of way is doing. Cause that's just using an ESP
7:08
chip inside a light bulb
7:11
or whatever it might be. And there was a Digi
7:13
Blur video just this week, which I'll put a link to
7:15
in the show notes as well, where he's talking about
7:17
an update to the liberation
7:20
script we've talked about on the show before, where you
7:22
act as
7:22
the man in the middle between the local
7:24
two year device and their update server and flash
7:27
ESP home onto it that way. There've
7:29
been some more updates to that recently as well.
7:31
So go ahead and check that out down below.
7:33
I've
7:34
been very, very tempted to look
7:37
at ESP solutions around
7:39
buttons
7:40
for home automation.
7:42
There's a lot of options, you know, Hue makes something,
7:45
of course, there's tons of Zigbee buttons, there's tons
7:47
of Z-Wave options. So I've been trying to figure
7:49
out, is that the route I wanna go if I wanna start
7:51
putting in more and more buttons to like turn on and off water
7:54
pumps and water heaters, or
7:56
do I wanna use something pre-made
7:57
that's maybe just on Zigbee or Z-Wave?
7:59
own some of these. So
8:02
I've been experimenting down that route this week and
8:04
I'm curious Alex, have you ever seen anything around
8:07
using an ESP home to just
8:08
essentially have like a button pad that you could
8:11
press just to and then just tie automations
8:13
and home assistant to when you press a button? Not
8:15
an ESP home device
8:18
but you could certainly create
8:20
a macro pad like a numpad
8:23
keyboard and put some kind of a microcontroller
8:25
on that and then use the matrix
8:27
layout of those keys to do
8:30
different things. People build all sorts of
8:33
stuff to go on like their smart desk setups. Obviously
8:35
the downside of those is typically they require
8:38
power all the time where an ESP device
8:41
depending on how clever you are can
8:43
potentially be battery based but the advantage
8:46
of using one of these sort of keypad
8:48
style things is
8:50
it's almost infinitely configurable
8:52
especially if you start delving into the world of
8:54
layers and all that kind of other stuff too.
8:57
So this is this is the problem. So the wife
8:59
she tells me you know I've been getting up early been
9:01
doing the things doing my things getting my coffees
9:04
doing a meditation and I don't feel like
9:06
using the tablet in the morning.
9:09
Okay all right and then I noticed that the
9:11
kids haven't been using the tablet like in the evening
9:13
like they just kind of it sort of slowed
9:15
down and
9:16
so I kind of felt like maybe I'd built a system that the family
9:18
wasn't really using and
9:20
kind of cut them out and I felt bad about them
9:22
I thought well okay I haven't really
9:24
used buttons because we mostly use voice
9:26
control and these tablets that are mounted
9:29
but maybe maybe
9:31
a button to control this specific light or
9:34
to kick off this particular automation or to control
9:36
this group maybe it's time for that.
9:38
So I decided to start doing some
9:41
digging in this area and I'm just not very impressed so
9:43
far. Zigbee buttons in
9:45
my experience kind of suck they
9:48
they work most of the time but
9:50
they failed just often enough
9:52
for you to be like did it work has it worked
9:55
oh yeah there it goes. Yep
9:56
that is it they
9:58
sleep right to save battery
9:59
Yeah. And some buttons do
10:02
support being plugged in, but then those buttons have
10:04
limitations. And then the tooling in Home
10:06
Assistant is pretty rough.
10:09
You can add a device easy enough using
10:11
Zigbee or Z-Wave or whatever you're using, Wi-Fi
10:13
if it's a Shelley potentially, but
10:16
then like, how do you do anything with that button?
10:19
Well, you have to go create an automation. Okay. Well,
10:21
now you have to figure out
10:23
what button the device thinks you're
10:25
pressing. When you press that, find the right option
10:28
in the automation, which sometimes has like 25 entries
10:30
in there, even though it's only got one or two
10:32
buttons. And then you have to create
10:34
an automation for everything you want each iteration of
10:37
those buttons to do. So if you've got like a quad
10:39
button panel, you have to create an automation for each
10:41
button. And then of course, all of these things
10:43
support, like if you double tap, it does something different.
10:45
If you hold it, it does something different. Well, that's a different
10:47
automation
10:48
for that.
10:49
And so if you got a few, a fair amount of lights or things
10:52
that you want to be able to
10:53
control over buttons, like I want to have a button,
10:55
the wife can hit
10:56
and it just raises the temperature five degrees.
10:58
Just
10:59
five degrees warmer for two hours.
11:01
And
11:03
there's just not a really great solution
11:05
and a home assistant for this. I have wanted
11:07
that kind of, uh, for
11:10
me, it would be cool. A house, you know, I'm, I'm
11:12
feeling hot right now and I will know I will
11:14
forget to turn the thermostat back up again. Can
11:17
I just have a, I'm playing the drums and I'm hot
11:19
right now button.
11:21
Yeah. Can that just be a thing? If anybody
11:23
knows how to do that kind of, I think it's probably
11:25
a scene. And then you return to the previous scene.
11:28
If you were in the audience and you have a working example
11:30
of that with code and the buttons, please
11:34
write in and let us know.
11:36
I started playing around with and I haven't gotten it working
11:38
yet, but I want to let the audience know because I think
11:41
this should be built in to home assistant. It's
11:43
called home assistant switch manager.
11:46
You can install it through hacks or do
11:47
whatever you like. And it gives
11:49
you a UI to set up your
11:53
buttons. And when
11:55
you add a device, it gives you, okay, here's
11:57
all the devices we know home assistant natively supports.
11:59
select one from the list. It's also just,
12:02
for God's sakes, nice to have at least some list
12:04
of devices that you know work with Home Assistant.
12:06
So that's another reason why this
12:08
plugin is nice. So you go through the list
12:10
of the buttons that work with Home Assistant, and then
12:13
once you select it, it has this brilliant
12:16
feature called auto detect. So
12:18
you put it in auto detect mode, and you press
12:20
a button on the switch, and it figures
12:23
out what button and switch and everything is, and then
12:25
you just start setting up graphically,
12:27
this button does this, this button does that, if I
12:29
press it twice, and it's all a nice UI,
12:32
and it doesn't require creating automations.
12:35
It is a custom integration, so you have to get that installed,
12:37
and then as a front end component you have to use, but it's
12:39
beautiful.
12:40
It's very minimally designed, it looks like something the Home Assistant
12:42
team might create.
12:44
The only problem with it is, is I can't get it to
12:46
work. It recognizes I press the buttons,
12:48
but then it doesn't execute the thing it's supposed
12:51
to do, like turn device on or off.
12:53
So I've had to bail on using it, but it also supports
12:55
MQTT for devices that use
12:57
that, and it supports Z-Wave and ZigBee devices,
13:00
it Bluetooth, anything that Home Assistant
13:02
can support it will work with.
13:04
I just haven't got it to actually execute the functions. I
13:06
imagine there's probably something wrong on my machine, so I'm back
13:08
to using automations, but I wanted to let you guys know
13:10
because this,
13:12
this is so good, it needs to be built in. They
13:14
need to build this into Home Assistant.
13:16
We know what Steve Jobs would say right now, don't you?
13:18
What? You're holding it wrong. Oh,
13:21
I thought he'd say something about don't have buttons or something.
13:25
I felt like I had discovered a game, you
13:30
know when you're like, you're gonna redo everything, you're gonna delete all
13:32
these automations, you're gonna redo the whole way
13:34
you did manage all this stuff.
13:36
I thought I was gonna do that with Switch Manager, but
13:38
I'm just not there yet. And so I'm just
13:40
stacking more automations.
13:42
I do have a couple of switches that have
13:44
worked for me
13:46
so far, and, but
13:48
like Alex said with the big caveat that
13:51
all of this stuff will like go to
13:53
sleep if it's battery powered.
13:55
You know the other thing that happens with those battery powered
13:57
buttons, if you're as lazy as I
13:59
am.
13:59
anyway is you go to push the button
14:02
one night and you think, ah well maybe it failed I'll
14:04
just get my phone out tonight and then you push it again
14:06
the next night and you're like
14:08
ah well it failed I'll fix it tomorrow
14:10
and then before you know it your buttons been out of battery for six
14:12
months and and your routine is completely
14:15
devoid of physical button presses and
14:17
this thing has been sat on the wall for six months
14:19
doing nothing so
14:22
yeah I just wish there was a way
14:25
an easier way for me to tap
14:27
into the always-on power
14:29
inside a light switch
14:31
to power some of these buttons like I
14:33
know there's that is it the zoos switches
14:36
I think we talked about and there's a wave
14:38
yep
14:39
and they go in the wall and you can wire them I
14:41
really want just some kind of a
14:43
non-offensive
14:45
button pad which looks like
14:48
a light switch and behaves like a light switch for
14:50
normal people and for me in
14:52
the middle of the night to be perfectly honest with you
14:54
but also has the smarts that if I want to
14:57
you know arm the front door
14:59
lock and do a bunch of stuff as I'm heading
15:01
to bed you know I don't have to pull my phone out every
15:03
time that
15:04
would be really nice if you know of anything
15:06
like that again please I see a crowdsourcing
15:09
episode in full today
15:11
but if you if you have any really good examples
15:14
like the zoo switches but Zigbee
15:16
would be my preference if you know any of anything like
15:18
that
15:19
please write in and let us know.
15:22
Linode.com slash SSH
15:24
that's where we host everything that we put
15:26
in the cloud anything that the listeners are going to
15:28
touch anything that we want to have super fast computer
15:30
infrastructure it's Linode
15:32
and they've got some exciting news to go
15:34
to Linode.com slash SSH get that $100
15:36
in 60-day credit you can really
15:38
kick the tires you can really try the infrastructure
15:41
and you can check out the great news.
15:44
Linode's now part of Akamai all the developer
15:46
friendly tools like their cloud manager which is beautifully
15:48
built
15:49
their API which is clean and well documented their command-line
15:51
client which is super handy all that stuff like
15:54
we use to deploy and scale in the cloud
15:56
that's still there it's still Linode but now they're
15:58
combined with the power and glory
15:59
reach of Akamai and they're expanding
16:02
the services to offer more resources and tooling
16:05
while maintaining that reliable affordable
16:07
and scalable solutions that open-source projects
16:09
individual listeners and businesses of all
16:12
sizes use and we love it
16:13
we run our matrix infrastructure in there and over the two
16:16
years we've scaled that thing into a monster to maintain
16:18
performance for our community and Linode handles
16:20
it the object storage is fantastic
16:23
I use it for all kinds of things now that are
16:25
just sort of like back in infrastructure bits and
16:27
pieces resources for clients
16:29
and whatnot for
16:30
accessing XML file pulling down a JPEG put
16:32
it on object storage Linode has
16:34
fantastic object storage and
16:36
as part of Akamai's global network of offerings
16:38
they're expanding the data centers worldwide they just
16:40
opened up a brand new one last week in Sweden it
16:43
is a banger and they're giving you access to even more
16:45
resources you can grow your business
16:47
project or your customers
16:50
so why wait go experience the power of Linode now
16:52
Akamai go to Linode.com slash SSH
16:55
that's where you go to get the $100 support the show
16:57
and learn how Linode now Akamai can
16:59
help you scale your applications from the cloud all
17:01
the way to the very edge
17:04
you know like Alaska probably
17:07
that's an edge
17:08
or Brent's house go try it support the show
17:10
and get that $100 to kick the tires Linode.com
17:12
slash SSH
17:16
Now before we get to an interview with Antonio who
17:18
is the lead developer of the merger FS
17:20
project I came across
17:22
a really interesting article on TechCrunch
17:24
earlier matter
17:27
not the matter you're probably thinking of no
17:29
no not the home automation matter the
17:31
other matter the app which
17:34
lets you read stuff later and transcribes
17:36
and reads it back to you matter have
17:39
added podcast transcription
17:41
support today this
17:43
is getting more and more popular I've been hearing
17:45
from listeners that already just transcribe our
17:47
shows and so another tool you
17:49
know in in that in that cap is nice I don't
17:52
I don't think I'm familiar with matter I actually when
17:55
you put this in the doc I thought you were talking about the
17:57
protocol well I mean it's
17:59
thing is it's not self hosted
18:01
at all. And the reason I mention it today is not
18:04
for our core listener base, of course, it's
18:06
actually from gonna make me get
18:08
off my ass and look at how we can transcribe
18:11
some of the JB show notes automatically
18:14
using some of the whisper tooling that's getting really
18:16
good these days. The tricky
18:18
part, and again, if you know a way around
18:20
this, right in, let me know. That's
18:23
the theme this episode, isn't it? If
18:25
you have a good way of doing the
18:28
diarization. So Alex said
18:30
this sentence and then Chris said that sentence.
18:33
I
18:33
think typically drew our editor does
18:35
a pretty good job of making it so we don't step on
18:37
each other, which isn't always the case with podcasts.
18:39
So the diarization part should be fairly
18:42
straightforward for the most part. If
18:45
you have a good way of doing that with the open
18:47
source whisper tooling that you you're using,
18:50
let us know. We'd love to build it into our release pipeline.
18:52
One thing we could do to make that simpler,
18:55
it would require some help from Drew, but I've
18:57
talked to him about this is supplying
19:00
host tracks. So there's a Chris track and an Alex
19:02
track.
19:03
And so you give that to the trends to whisper
19:05
and whisper knows everything said on this track was from
19:07
Chris. Everything said on this track was from Alex. Oh,
19:10
yeah, of course we have that option is the
19:13
the creators of the content. Yeah,
19:15
we got it. And we have the source we got it multi track.
19:18
So we that's, that's what's cool about
19:20
being able to build it in at the
19:21
at the production levels, we could potentially do that. Yeah,
19:24
I'm really, really excited about
19:26
getting that rolled out because I've been experimenting manually
19:28
generating them manually attaching them to the shows for
19:30
some things every now and then trying different formats.
19:33
So close, like it still messes up on some
19:35
of the tech terms.
19:37
So like this doesn't seem totally feasible,
19:39
but a dream of mine would be
19:41
transcription gets published.
19:43
And then people could do poll requests against
19:45
it and the community could maybe fix the transcription
19:47
if they cared because it's not something we're going to go
19:49
back and clean up. You know, I could see
19:51
us running a better transcription five years down
19:54
the road and just overriding all of the transcription
19:56
files with a better version. I could definitely
19:58
see that at some point.
20:00
Alex, you had a chance to sit down with the lead developer
20:03
of MergerFS and he joined us for
20:05
a chat.
20:05
So welcome back to the show, Antonio. The last time we
20:08
spoke to you technically wasn't in
20:10
self-hosted. It was in one of the Jupiter
20:12
extras shows where we interviewed you with
20:14
Drew and Brent to talk about
20:16
MergerFS. How are you? I'm
20:19
well, how are you doing? Doing good. Thank
20:21
you. How is the new Texas sunshine
20:24
treating you? It's been hot.
20:27
It rained once in two months for like 15
20:29
minutes. We kind of celebrated. So
20:32
it's been a little bit to acculimate
20:34
from New York City,
20:35
but it's been good. Yeah, I tell you,
20:37
we were just talking about this before we
20:40
pressed record. I think if I lived
20:42
in Austin, I would probably have gained another
20:44
hundred pounds since moving
20:46
to America. Just some of the barbecue
20:48
down there is just next level.
20:51
The food in general, right? The Mexican food, of course,
20:53
and everything. They've
20:55
got all these fusion. My
20:58
wife and I, it was our ninth
21:00
anniversary of dating and
21:02
we went to this place. It's like
21:05
brisket, but with Asian flair
21:08
and really good. Came
21:10
home completely stuffedicated, fell
21:12
asleep on the couch afterwards. Delicious.
21:16
The fact that I work from home and I'm in my office
21:19
most days, five
21:21
days a week at least, yeah, I'm
21:24
lucky that I haven't put on that hundred pounds you
21:26
mentioned. That sounds amazing.
21:29
Now for those of you in the audience that don't know Antonio,
21:31
I'll forgive you because it was 2019, the
21:34
last time I think we spoke. Antonio
21:36
is the guy behind MurderFS. I
21:38
think one of the most underrated,
21:40
dare I say, like low
21:42
key, awesome bits of
21:44
Linux technology that's really changed
21:47
the way in which I interact with hard
21:49
drives in media servers in particular.
21:52
So the idea behind it is you have just a bunch
21:54
of drives
21:55
and then you point MurderFS at
21:57
them with an FS tab mount.
22:00
entry or something, and then it kind
22:02
of pulls those drives together into one
22:04
big, what
22:05
would you call it, like gluttonous mount point,
22:08
and then you can kind of traverse all of
22:10
the files and folders on
22:12
those drives as part of that JBOD,
22:14
as if it was just one single massive
22:17
drive.
22:18
With conditions, but yes. I
22:20
mean, if people are familiar with Drivepool
22:22
on Windows or UnionFS
22:25
on Linux or AUFS,
22:27
there's a few different technologies over the years. Union
22:30
file systems have been around for 30 years,
22:33
at least. If you use Docker or containers,
22:35
you might be familiar with overlayFS,
22:38
which is a different kind of Union file
22:40
system. So yeah,
22:41
I'm in that category.
22:43
All right, so let me ask you this. It's been
22:46
two or three years since we spoke, 2019, I
22:48
think, the last time. This is your opportunity
22:51
to tell the good people of the self-hosted podcast
22:53
what's changed. There's been a lot of random
22:55
stuff. I mean, the core features are all there. For
22:59
the average user, I don't think much has
23:01
changed, if anything, at
23:03
least from their perspective. Under
23:05
the covers, I've done a lot of cleanup. I
23:08
years ago embedded libfuse into
23:10
the project to make it easier for me to extend things.
23:13
And I did a lot of kind of retrofitting
23:16
of the code there. I reduced memory
23:18
footprint quite a bit and introduced
23:21
some techniques to just limit
23:23
fragmentation of memory, which was an
23:25
issue for some users. If you had a machine running
23:27
for a long time, there's a lot of churn
23:30
of objects, especially on SPCs,
23:33
smaller RAM systems that would cause issues.
23:35
So I've helped mitigate the amount
23:37
of memory in general used and then kind
23:40
of limited
23:41
that memory leak in the form of fragmentation.
23:44
I've also added, again, under the covers,
23:47
threading pools to certain behaviors.
23:50
This is one of these things as a software developer. You
23:52
have an idea of who's going to use your software in
23:54
a certain way, and then you release it on the world
23:57
and people use and abuse it for all
23:59
kinds of other purposes.
23:59
And I've had
24:02
folks with four socketed
24:05
Xeon systems running
24:07
it against local
24:10
shares, running different file systems,
24:12
connecting to remote file systems
24:14
of all sorts. And in those
24:17
situations,
24:18
the concurrency
24:21
can be both good
24:23
and bad. Good in that you've
24:25
got more things happening in parallel. But
24:27
because of
24:29
how scheduling works, it can
24:31
actually reduce the throughput. And
24:33
so there's features for pinning
24:36
threads on the cores to separate
24:39
receiving messages from the kernel with actually
24:41
processing them, allowing you to
24:43
determine how many readers you want, how many
24:46
processors you want, and then different
24:48
strategies for pinning the cores. And
24:50
that helps increase throughput.
24:52
I haven't released this yet, but soon I'm releasing
24:54
a feature that's been asked for for a while,
24:57
read dir, right? When you actually scan
24:59
directories, a lot of people will have
25:02
network file systems and the latency to connect
25:04
to those is pretty high, or
25:06
they have maybe spinning disks where they're
25:08
asleep.
25:09
And so I'm
25:11
concurrently connecting to or doing
25:14
a read dir on all of those at the same time,
25:16
if optionally, because it increases
25:18
the memory usage a bit. But that way it can
25:20
reduce
25:21
the latency to actually get that
25:23
data. And so when you do an LS,
25:26
it's faster. Yeah, I noticed Wendell
25:28
did a video fairly recently on ZFS
25:31
where he put his metadata onto
25:33
a pair of NVMe drives.
25:36
And even though the data was still stored on spinning
25:38
drives underneath, the lookup times,
25:40
the seat times, for like just listing the
25:43
contents of a specific directory was 10
25:46
or 20 times faster, just simply by
25:48
moving that metadata. Is that the kind of thing you're
25:50
talking about here? No, though, I
25:52
have been working on something similar. Merger
25:55
FS and a lot of union file systems at their core
25:57
is almost like it's just a union.
25:59
in the truest sense. Imagine you have
26:02
A, B, and C, and
26:04
you were to LS in each one of them individually.
26:07
Under the covers, that's all merger FS is doing. And
26:10
so imagine you type LS
26:12
and your drive has to spin up and it takes like 10 seconds.
26:15
Well, if each one is asleep and it takes 10 seconds
26:17
each, it's going to take 30 seconds at least in
26:20
aggregate. Now, what this
26:22
feature does is just issue each of those
26:25
at the same time and then aggregates
26:28
the data as soon as it's available. So you're looking
26:30
at more like 11 seconds rather than 31 seconds.
26:33
That sounds fantastic. And I guess, do you
26:36
have any sense of what the typical
26:39
merger FS deployment size
26:41
is? I mean, there's no telemetry or anything like that in
26:43
your packages are there. So it must be tricky.
26:46
It's one of these things where the
26:48
exceptions probably indicate
26:50
the rule. And what I mean by that
26:53
is the most questions
26:55
I get about merger FS are
26:57
usually from total noobs who don't
26:59
know anything about file systems. And so unfortunately,
27:02
there's only so much I can do to simplify
27:04
what a file system is. That's most
27:06
of us by the way, dude. You know that, right? Well,
27:09
yeah. But I mean, people who have like zero Linux
27:11
experience are coming straight from Windows. They really
27:13
have no
27:15
understanding of how file systems work even
27:18
from just a general purpose
27:20
user perspective. And so I get
27:23
a lot of questions from that.
27:24
But then the other side is
27:26
the people who have these, like I was saying earlier,
27:29
these crazy setups of like
27:31
multi socket Xeon systems with 100
27:34
threads or something.
27:35
I think there's probably a very large silent
27:38
minority or majority there of
27:40
people who just are, you
27:43
know, they've got five drives and
27:45
that that's kind of their setup. And maybe
27:48
they want an SSD in there on occasion. Well,
27:50
speaking of SSDs, I actually had a question for you
27:52
around caching. This is something on
27:54
perfect media server that I actually get quite
27:57
a bit as a question is it's
27:59
pretty common
27:59
in the un-raid world because of how
28:02
they do their parity calculations, you're
28:04
basically halving your write speed of any disk
28:07
because for every single write you make,
28:09
it has to make another write to the parity drive
28:11
as well. So it basically just cuts your write performance in
28:14
half, which is why un-raid many
28:16
years ago adopted that cache drive
28:18
and then mover script type stuff. Now
28:21
I know there's some stuff in your read me about, is it B
28:23
cache I think and a bunch of other stuff.
28:26
What's your take on
28:28
caching? So there's lots of levels
28:30
of caching and this can be
28:32
very confusing for folks. And unfortunately,
28:35
again, it's one of these things where
28:38
the features are there for functionality
28:41
purposes. If there was one great generic
28:43
way to set it up, I would just make that the default.
28:46
Unfortunately, I find that that's not the case,
28:48
especially since a lot of people are using
28:50
it in a way
28:52
where if you did induce caching,
28:55
people want to write things out of that band. They want to be
28:57
able to write to our clone independently
29:00
and still have merger FS work. And you
29:03
can't have caching there because there's no way
29:06
you'll eventually get into a bad state with that.
29:08
So there's certain, there's kind of caching
29:10
in the kernel and merger FS has
29:12
a number of features there
29:14
that are related to fuse directly. Then
29:17
you have
29:18
caching that merger
29:20
FS itself can do. I
29:22
don't do too much of that, though I'm looking
29:24
at doing some more of it just to reduce the
29:27
amount of calls I have to make into the kernel.
29:30
And then there is usually what they call
29:32
tiered caching for the underlying
29:34
devices. And that is where
29:37
you have NVMe or Optane
29:39
in front of spinning disk or Optane in front
29:41
of SSD in front of spinning disk,
29:44
that tier of setup. And
29:46
this is something that I hope to fix in the next year.
29:49
Because merger FS has kind of a simple
29:52
key value pair, config setup,
29:54
it's difficult to articulate to the software
29:58
something like a
30:00
very thorough tiered caching
30:03
system. But there are ways to implement that
30:06
regardless out of band. And that's the kind
30:08
of things I generally show in my documentation,
30:10
which is similar to those mover
30:13
scripts that Drivepool has or
30:15
Unrate has. What you can do is just
30:17
create two pools
30:19
and you can
30:20
set one pool up as your primary
30:22
pool. And that's where you put your SSDs, your fast
30:24
storage and your slow storage. And the idea
30:26
is that you create a policy, that you have a policy
30:29
in mergerfs. Every
30:31
kind of file system function has a policy.
30:34
And that policy is what
30:35
chooses how to behave when
30:37
that function is called. So for instance,
30:39
you want to open a file.
30:41
There's a policy that gets run. And it
30:43
chooses which file is going
30:45
to be opened. Because you could imagine a scenario
30:47
where you have five drives and you have four files across
30:50
five drives, like how do you pick one?
30:52
What you do is you set up a creation policy
30:55
such that your SSDs are prioritized.
30:57
So mergerfs is kind of always picking your
31:00
SSDs. Then you create a secondary
31:02
pool.
31:03
And the reason you want the secondary pools, mostly
31:06
because it's easier than duplicating
31:08
the logic of moving stuff
31:10
from drive to drive. But the idea is you
31:12
just
31:13
then target, like with our sync,
31:16
that SSD. And you just, every
31:18
day or whatever, you just move those files
31:21
over to the secondary pool. And that
31:23
secondary pool has none of the SSDs.
31:25
It only has your slow devices in it.
31:27
And then so far as something like Plex
31:30
or Jellyfern is concerned, the
31:32
files haven't moved. They're
31:35
still in the same place, most likely,
31:37
because of how the mergerfs union
31:40
stuff works. There is some subtlety there,
31:42
but yes. Yes, and it's a neat idea. And
31:44
I think you could very easily combine
31:47
that approach. I mean, that's how I've been adding
31:49
ZFS into my single
31:52
storage mount point for the last few years.
31:55
It's a really interesting concept. And I think once
31:57
you unlock that idea, that you could have more than one.
32:00
one merger FS mount point on a single system
32:02
that's containing different tiers
32:05
of storage, for one of the better phrase, different classes
32:08
of storage, then
32:11
you can ramp up the complexity quite quickly,
32:13
but it also ramps up the flexibility
32:16
massively. Well,
32:17
yeah. And if you think about it, so that
32:19
idea to add SSDs in that
32:22
form came after
32:24
a very popular usage of merger FS, which
32:27
I didn't see it coming when I first
32:29
created it, which is people will have a local
32:32
cache of drives,
32:33
and then they have their
32:34
data hoarder, all
32:36
their ISOs sitting on a
32:38
Google Drive or something, and they tend to use RClone.
32:41
And the author of RClone and I are on
32:44
good terms. He has a union feature
32:46
in RClone that is
32:48
mimicking what merger FS does. And
32:51
so people will combine
32:53
the two and basically use their hard drives
32:56
as a cache to cloud storage. And
32:58
so you just add in SSDs in
33:00
front of all that and you have,
33:03
again, another tier. There's another
33:05
strategy that can be done
33:07
that I don't see a lot of people doing. I created
33:09
a tool years ago to orchestrate all
33:11
this stuff, but I never
33:13
made it
33:14
popular, I never promoted it. And I've
33:16
been looking at maybe doing that again. So if you're
33:19
familiar with
33:20
device mapper in Linux, a DM
33:22
setup, which is used, it's the underpinning
33:25
of technologies like LVM2,
33:28
you can actually take a block device,
33:31
any random block device, whether
33:33
it's a hard drive or SSD, and you can make
33:35
another device a
33:38
block cache for it. There's
33:39
DM cache, which a lot of people
33:42
know. And so you can use LVM to
33:44
create logical volumes and you can create
33:46
a cache partition. But what a lot
33:48
of people don't realize is you can actually take a random
33:50
hard drive you have already formatted with whatever
33:53
file system you have, and use
33:55
an SSD to create a cache on top of
33:57
that.
33:58
It's not that it's some secret issue. just not well,
34:01
it's not
34:02
publicized, I think, because there's no infrastructure
34:04
around it, no software to kind of automate that
34:06
process to say like, okay, I want,
34:09
you know, 100 gigs of this SSD
34:12
to be cached for this drive and 100 for
34:14
that and, and to be able to easily bring
34:16
it all together. And so I've been looking at, at maybe
34:18
releasing a tool that can help with that as well.
34:21
And so that way, people who want that caching
34:23
at a block level, which can help with spin
34:25
up of drives and other issues can
34:28
do that without having to move to ZFS
34:31
or move to B cache or B cache
34:33
FS.
34:34
We'll see that where that goes. Mason
34:36
W
34:40
Most
34:43
of the changes or additions that I've made in the past year or so have
34:45
been quality of life things. So
34:47
I added like O Direct Support, which is
34:49
kind of a niche feature that some programs
34:52
use, but it took a little bit to
34:54
get working. So I added that I added the ability
34:57
that if a drive goes
34:59
into read only mode, it will tag
35:01
it as read only. And so it'll find another
35:03
drive, right? Like you try to create a file and it's a read
35:05
only device because your extension
35:08
for partition got corrupted, it'll
35:10
find it another drive that will
35:12
work if one's available. Read
35:14
ahead, setting read ahead has always been kind
35:16
of a pain. So merger FS will do that
35:19
for you.
35:20
The ability to do lazy unmounting of
35:22
an existing mount point,
35:24
because fuse doesn't have a good way to remount like
35:27
traditional file systems do merger FS
35:29
can take care of that for you. And so as soon
35:31
as the last program stops using the old
35:33
version of merger FS, it'll just get unmounted.
35:36
So there's been a lot of stuff like that.
35:38
Going forward, what I really want to do is
35:40
completely redo the configuration system
35:43
and move to Tomel
35:44
for the config. And what that
35:46
will enable me to do is add
35:49
a lot of features that people have been asking for
35:51
in terms of like built in tiering
35:53
knowledge so people would really like merger
35:55
FS to know
35:57
about the tiers of you know, the
35:59
performance, difference, performance characteristics of different
36:01
drives. So it can more
36:03
intelligently choose which file
36:05
systems choose
36:07
when it's creating a file or reading a file
36:09
or whatnot. And that's kind of a big lift.
36:11
It might not sound like much, but because the
36:13
configuration is
36:16
very simplistic at the moment, and
36:18
it touches a lot of pieces, right? Like there's a runtime,
36:21
config option, and all these things. So that's
36:23
going to take a little bit of work to do, but
36:25
that's going to enable me to
36:27
add probably like a list of instead
36:30
of just having a list of branches, you can
36:32
have like a list of lists
36:35
of branches, right? So you could have a
36:37
collection of SSDs or collection of NVMe's
36:40
that get priority and then that falls back if
36:42
nothing fits there into the slower
36:45
drives. And so that'll allow for
36:47
that tiered caching that we talked about in a
36:49
more fluid and more natural way.
36:52
I also have
36:53
been thinking of
36:56
adding features kind of like what Unraid
36:58
and Drypool do. I
36:59
think the reason that they have
37:01
mover scripts is because they want to have
37:04
a very discrete, simple
37:06
way of laying out the
37:09
files, right? As you create files and
37:11
whatnot. And then they worry about moving
37:13
them on after the fact,
37:15
kind of like the mover script that we talked about earlier.
37:17
A lot of users
37:21
are either because they've used those products,
37:23
I think, or maybe it just seems the
37:25
more natural way of doing it would really like the
37:27
behavior of like rebalancing, right? Like I
37:29
add a new drive and I want it to kind of slowly
37:33
move stuff around in some fashion.
37:35
And merger FS doesn't do that very explicitly.
37:37
It chooses a position where
37:40
it's going to create something, you know, on the fly as it's
37:42
making a decision, and you get to choose that policy.
37:45
And so I think it'd be better to have both, right?
37:47
Like you can choose up front, but then there's kind of a
37:50
background task that'll just sit and move things
37:52
around more subtly. So
37:54
that's a big one that I'm going to look at as well. I
37:57
do wonder sometimes just how
37:59
much of this is started.
37:59
syndrome, a lot of Unraid users
38:02
in particular. No disrespect
38:04
to that project, because it's been around forever. Lots of people
38:06
use it. Lots of people like it. But it does
38:08
do some stuff in a strange way. And
38:10
it has to do things in a certain way to
38:13
make up for decisions made 15 years ago, booting
38:16
from USB being a great example. So
38:19
anyway, I wanted to say thank you very much for coming on today, Antonio.
38:22
I really appreciate it. Where can we send
38:24
folks if they want to support the project, or open
38:26
a feature, or a bug request, or something like that? So
38:29
if you just go to
38:29
your favorite search engine and type MergerFS,
38:32
it will show up. The main page is
38:35
just the GitHub page. So it's github.com
38:37
slash trap exit slash MergerFS.
38:40
And if you're interested in supporting the project,
38:43
instead of MergerFS, go to just support. So there's
38:45
actually a repository that has all the support details
38:47
there. Lovely. Yeah, all the GitHub sponsors,
38:49
Open Collective, Patreon, all that kind
38:51
of good stuff. Thanks so much for coming on,
38:54
and enjoy the rest of your day. Thanks for having me.
38:58
Tailscale.com slash self-hosted.
39:00
Head on over there right now to get Tailscale
39:02
for free for up to 100 devices. It's
39:04
a great way to support the show.
39:06
And you can get the zero config VPN up and going
39:08
in just a couple of minutes, and have a mesh network
39:10
between all your devices, protected
39:13
by WireGuard's noise protocol.
39:15
And it's fast. It's really fast, because
39:17
all your machines talk directly to each other.
39:20
And the Tailscale client is smart enough to know if you're trying
39:22
to send something to a machine that's destined
39:25
for your tail net, like something that has
39:27
a tail net IP, or if it's something for the internet.
39:29
And
39:30
that small little difference makes a
39:32
big user impact, because you
39:34
can leave Tailscale running all the time. Like I have it on
39:36
my Pixel 7, 24-7, always connected.
39:40
But only the resources I need on my tail net go over my
39:43
Tailscale connection. So I put all of my infrastructure
39:46
on my tail net, all my devices,
39:48
my NextCloud, anything I'm going to sync to, anything
39:50
that I might, like my pictures backup, Home
39:52
Assistant, I put it all on Tailscale.
39:55
So that traffic goes over Tailscale.
39:58
And then there's a lot of nice to have, like Tailscale.
39:59
SSH which lets you log into any machine
40:02
on your tail net using your tail scale credentials
40:04
and ACLs
40:06
And yeah, they got a dashboard to let you manage all that. There's
40:08
things like tail scale send
40:10
Tail drop, you know kind of like airdrop,
40:13
right? But it's
40:14
it lets you send files between your tail scale
40:16
machines They have a plug-in for VS code
40:18
so you can edit the config files on any machine
40:20
in your tail net They have lots of ways to extend
40:22
it
40:23
Alex and I share one machine between each other and
40:25
you can even limit it to the ports That
40:27
you can allow through there. They have a
40:30
client for just about every architecture and OS
40:32
mobile desktop server SBC
40:34
but
40:35
tail scale also supports something called
40:37
subnet routing. So like my solar equipment
40:39
I can access that even though I can't put
40:41
the tail scale client on my solar equipment
40:44
I have a machine on my network
40:46
where I have subnet relay turned on and so using
40:49
that I can get to those systems for My other tail scale
40:51
clients and I love that
40:53
So I always know how the house is doing even
40:55
the electrical equipment
40:56
Go try it out support the show and get it free for up
40:58
to 100 devices when you go to tail
41:00
scale comm slash self
41:02
hosted
41:04
Well,
41:04
you've been asking and I think we
41:07
have an answer for you
41:09
It's always the llamas llama GPT.
41:11
L. L. A. M a GPT a
41:13
self hosted offline
41:15
Chat GPT like chat box powered by
41:18
llama to
41:19
no data leaves your device and
41:21
They just added code llama,
41:23
which is one of the later models that Facebook just put
41:25
out and they just added support for Nvidia
41:28
GPUs
41:29
It's created by the umbral folks
41:31
and they've released this just as a general
41:34
document so you can put it on any system
41:36
Some of this stuff is getting crazy good just
41:38
to go back to the transcription stuff for a second There
41:41
is a version of whisper for Mac called
41:43
Mac whisper which I've been using to transcribe
41:46
all of my YouTube content recently
41:49
and
41:50
I feel like having a local chat GPT
41:52
like bot, you know where I could maybe
41:54
feed that You know that transcription
41:56
of a clip. I've just recorded and say
41:58
could you?
41:59
maybe make that snappier or just, you
42:02
know, some of those little things that you use chat
42:04
GPT for. It's always at the back of my mind
42:06
when I log into OpenAI's website of,
42:09
where is this data going? So
42:11
I love, I love, love, love that this is 100% private
42:13
and local. And
42:16
the UI is beautiful. They really picked like
42:18
the best
42:19
of the front end software that's out there right
42:21
now and combined it with the latest
42:23
and freshest of the open source large language
42:26
models
42:27
all on Docker. And they worked really
42:29
hard
42:30
to make it, although I wouldn't do it, but they made
42:32
it, they made it possible to run it on
42:34
a Raspberry Pi.
42:36
I mean, it's horrible performance.
42:38
Like for example, if you're on an M1
42:41
Mac, MacBook, you're going to get a
42:43
generation speed of 54 tokens a second
42:46
on a Raspberry Pi 4 with eight gigs of RAM,
42:48
you're going to get 0.9 tokens a second.
42:50
Okay. So it's a big difference. Why
42:53
is anybody still running a Raspberry Pi 4 at
42:55
this stage? You know, maybe you're like me and
42:57
you're an old man and you're like, I like it slow
43:00
sometimes. I know that sounds weird, but it's
43:02
nostalgic when the computer's slow. I'm
43:04
being facetious. Of course I am. I understand.
43:06
There's plenty of good reasons.
43:08
It is really nice to have it all local though. I agree with you, Alex
43:11
and to have code llama local too is choice.
43:14
Llama llama duck. This
43:18
speaks to a piece of work I've been doing with, do you remember
43:20
Morgan, the doorbell guy? Yeah. He
43:23
and I have been working on a quick sync benchmarking
43:26
script over the weekend. Finally,
43:29
finally, finally, we wanted to work on this for a long
43:31
time. I think I mentioned it in the show a few episodes
43:33
ago as well. But essentially
43:35
I edited together some of my old drone footage
43:38
into a two minute clip of the
43:40
Ribblehead Viaduct in the UK from a few years
43:42
ago. Uh, not the one that's used
43:44
in Harry Potter. This one's up on Blea Moore
43:47
in the Yorkshire Dales. Beautiful, beautiful
43:49
structure. Anyway, I digress. The
43:52
purpose of this script is to try
43:54
and get a sense for where the sweet spot
43:56
is with quick sync.
43:58
Uh, I had... itchy feet the other
44:00
day and I thought I wish I had
44:02
a server with more PCIe lanes
44:05
so I could put some more NVMe storage in this thing.
44:08
But my trusty i5-8500
44:10
CPU and the motherboard combo
44:13
I have in there which is an ASRock rack motherboard,
44:17
they don't really have much in
44:19
the way of PCIe lanes so I was thinking
44:21
well could I go Ryzen? Is
44:23
that a thing? What about an ARK GPU?
44:26
Could I use that? But then isn't that going to quadruple
44:29
my energy usage? I thought
44:31
to myself I don't know. I don't want to buy these
44:33
things and find out later. So what
44:35
I'd love to do and it might be ready
44:37
by the time we record the next episode
44:40
but keep an eye on the Discord server, there's an active
44:42
thread over there
44:43
on the Perfect Media Server channel called QSV
44:46
testing, quick sync video testing.
44:49
Where we're talking about the various
44:51
different iterations of this script and how we can
44:53
do benchmarking across all
44:56
or as many of the different quick sync
44:58
encoding engines as we can. I've
45:00
got access here to second,
45:03
third, fourth, I think sixth
45:05
maybe and eighth gen Intel
45:07
CPUs and a ninth actually. So
45:10
if you have something else or indeed you
45:12
have one of those and you'd like to run the test as well,
45:15
join the Discord and let me know in that channel
45:17
and we'll share the GitHub repo with you where the script
45:20
is. The idea is to try and figure out where
45:22
the sweet spot is in terms of price
45:24
to performance, in terms of codec
45:27
support, all that kind of stuff. Why hasn't this
45:29
been done already Alex? Why hasn't this
45:31
been done already? I don't know. I mean
45:33
I look at what LTT labs are up to and
45:35
you know gamers Nexus and all that and they're all focused
45:38
on gaming. Like
45:39
that's fine but the terminal
45:41
is my favorite video game. I don't need a 3090
45:44
you know. The hardware can do other things besides
45:47
play video games you guys. Yeah
45:49
sometimes. Yeah
45:50
sometimes. That'll be really
45:52
great to see the results. I'm very curious to see
45:54
how that plays out. The quick sync sweet spot.
45:57
That's gonna be awesome. Yeah
45:58
well when the benchmarks are all right.
45:59
and all the rest of it, I'll make a proper
46:02
blog post, perfect media server page,
46:04
podcast episode, YouTube video, the
46:07
virtuous cycle of content will be strong with
46:09
this one because it's a lot of effort. I am very
46:12
tempted by the ARC GPU. I have
46:14
one in a machine in front of me and
46:16
it sings with Linux, everything
46:19
works so smooth, it's so snappy,
46:21
full Wayland support, it's all
46:23
just flawless.
46:25
However, I constantly
46:27
struggle with tools like stable diffusion
46:29
or Lama GPT or even
46:32
video encoders, they just don't
46:34
even grok what the Intel ARC is.
46:37
Nobody has really built support and then you have to go
46:39
find, as far as this is my understanding, like
46:41
if you want to run something like
46:44
say Lama GPT or stable diffusion using
46:46
the Intel ARC, you basically got to go get patched version
46:49
of the project. Well, you could do what
46:51
Wimpy does and run multiple
46:53
GPUs in the same system and just have a headless
46:55
Nvidia card that has the CUDA
46:58
driver available for those particular apps.
47:00
Not an awful solution if you've got the
47:02
card already, like if you've got the hardware, I don't think
47:05
I'd go out and blow the money on a high-end Nvidia, not
47:08
to use it for anything else, but if I already had
47:10
one.
47:11
Yeah, also not only do you have to have the physical
47:13
hardware, you've got to have the minerals to configure it too,
47:15
it's not a simple task.
47:19
45homelab.com, big,
47:21
strong, fast storage servers with affordable, high-performance,
47:24
high-capacity enterprise storage solutions
47:26
for
47:26
all industries, for
47:28
all data size requirements. I mean, we're talking
47:30
professional-grade solutions that
47:32
are ideal for a business, maybe your
47:34
home lab too. So go check out
47:37
45drives.com to learn more about those folks. You might remember
47:39
them from the show before and
47:41
they have been cooking up 45homelab.com.
47:45
What if you took all the ideas and all the
47:47
skills learned and all of the workmanship
47:49
that went into the enterprise-grade storage,
47:52
but you made something just for the home
47:54
lab? That's their mission. They want to change
47:57
the storage market and they think
47:59
they have a vision.
47:59
for the future Homelab product market.
48:02
They've
48:02
been listening to feedback from our audience, because we had
48:04
them
48:05
back in self-hosted 98,
48:07
and they're cooking up 45homelab.com. I
48:09
think that's going to be up your alley, so go check that out. Again,
48:12
it's 45homelab.com
48:14
for the stuff they're working on for us, Homelabers.
48:17
And I think 45 Drives maintains
48:19
probably one of the best relationships with
48:22
the open source community. They have open designs for
48:24
their hardware.
48:25
I really like their overall ethos
48:28
and where they take this stuff. So I think you might like it too.
48:31
So go learn how 45 Drives does things differently
48:33
at 45drives.com.
48:35
If you get a chance, tell them the self-hosted show sent you if
48:37
you end up buying something.
48:39
And take a minute and go
48:41
visit 45homelab.com. I
48:43
know it's two URLs. It's tricky. One's
48:45
the company, 45drives.com,
48:47
and one is the project they're cooking up for us self-hosters,
48:50
45homelab.com.
48:54
We got a doozy of an email into the show
48:56
this week. I think it's Keone's how you pronounce it.
48:59
And he lost his Homelab
49:01
in a fire. In fact, his whole town
49:03
went
49:03
up in a fire. And he doesn't have
49:05
a lot of budget to work with. He's got some networking
49:08
limitations,
49:09
but he's rebuilding.
49:11
And he's picked up an older HP ProDesk
49:13
for 20 bucks, an i3, fourth gen,
49:15
500 gigs
49:17
of hard drive and four gigabytes of RAM,
49:19
which he can upgrade over time.
49:20
And so now he's trying to wrap his head around how to kind of restore.
49:23
And he says, what would be the best and easiest
49:26
and
49:27
probably the most transferable way to get my old
49:29
systems up and running? I'm staying with family,
49:31
so I don't currently want to mess up with any of their router
49:33
settings or their firewall.
49:35
I'm wondering if TailSky or WireGuard could help here. Well,
49:37
first of all, congratulations on finding that HP ProDesk
49:39
for $20. That's a bit of a steal, isn't it?
49:42
Really? Yeah. Well, I'm very much currently
49:44
in love with those small and cheap one liter PCs.
49:46
You know, the the one liter style that you can
49:48
find refurbs on Dell's website or
49:50
eBay for sort of 100 ish dollars. More
49:54
modern stuff might be better from a power
49:56
draw perspective. So like the eighth gen
49:58
draw seven watts idle.
49:59
whereas the fourth gen draws closer
50:02
to 20. I don't know if that's a consideration
50:04
for you, but I seem to recall
50:06
power on Hawaii isn't the cheapest
50:08
thing in the world. The
50:10
other thing to consider is that you could potentially just
50:12
use something quite turnkey,
50:15
like dare I say, tail scale, to
50:17
connect into these remote devices without having
50:19
to do a whole bunch of firewall punching and configuration
50:22
and stuff.
50:23
Naked Wireguard is great if you
50:25
have access to the firewalls and you're
50:27
familiar with distributing your own
50:30
keys and stuff like that, but I
50:32
don't know what your appetite for that kind of thing is.
50:34
Yeah, I think tail scale is probably the way to go on this one too.
50:37
And then you're not messing with anybody's network.
50:40
If you move it, they'll reconnect to each
50:42
other, reestablish. If you end up on a different network down
50:44
the road, you're not going to have to rebuild your
50:46
VPN setup at that point, which you'll probably become
50:49
pretty familiar with.
50:50
He continues to say, I'd like to have a Nextcloud
50:52
instance using an external USB one terabyte
50:54
drive, piehole for ad blocking,
50:57
maybe for my devices only on tail scale, possible
51:00
sombar NAS drive for file sharing, and
51:02
Plex or Jellyfin.
51:04
I mean, that's kind of the self hosted recommended
51:06
setup there. Piehole is pretty great, Plex or
51:08
Jellyfin, whichever one fits your use cases, I'd
51:10
say start with Jellyfin.
51:12
And then if you have issues, try Plex.
51:15
Also, he says he's considering Proxmox.
51:17
I think that's a good idea. He says I'm okay messing
51:20
with any type of install in a base Ubuntu server Proxmox
51:22
or Docker.
51:23
Now I don't know about it. What do you think Alex? Proxmox
51:25
on an i4 with four gigs of RAM.
51:28
That might
51:29
be tight. I mean, you're going to run up against
51:31
the limits of four gigs of RAM pretty
51:34
quickly, as soon as you spin up one
51:36
virtual machine. But
51:38
if you were to use Proxmox to manage a couple of LXC
51:40
containers, then that gives you a
51:43
lot more runway, of course, and Docker
51:45
containers as well. Of course, Proxmox is just Debian
51:47
Linux under the hood, so you can do that too.
51:49
I get lots of people asking me
51:52
how I run my perfect media server
51:54
setup. I updated the FAQs today
51:56
with the answer to the question of should
51:59
I run perfect media? server on the host directly
52:01
or as a VM? Because for some reason that's
52:03
a really important question to folks, I've never
52:05
really quite understood. No, I don't
52:07
need to shit on people that way. The answer is,
52:10
it's really up to you. If you want to
52:12
run some services in a VM and have that
52:14
level of isolation,
52:15
that's great. You're gonna have to do a couple of things like
52:18
pass through if you wanna have disks available, that kind
52:20
of stuff. Or set up file
52:22
sharing from the Proxmox host into
52:25
the virtual machine using maybe
52:27
some kind of internal bridging or
52:29
something like that. Vert FS or Vert
52:32
9P I think is what it's called for Windows
52:34
hosts.
52:35
A lot of people like to keep the Hypervisor host
52:37
clean, but for me the trade-off of
52:40
running everything directly on the host because of
52:42
access to things like Quicksync is
52:44
kinda worth it. So,
52:46
you know, if you just wanna keep things simple, just
52:49
stick Proxmox on there. You may never
52:52
use any of the Proxmox features, any
52:54
of the virtualization stuff,
52:56
but if you do decide further down the
52:58
road that you wanna do that kind of stuff, you don't have to then
53:00
completely wipe your entire OS and start
53:02
from scratch.
53:04
Yeah, well said.
53:05
Dimitri is struggling to ditch iOS.
53:07
He says, I've been trying to switch from iOS
53:09
device to a Pixel
53:11
running GrapheneOS.
53:13
I've tried it three times, but
53:15
I've been using iOS for so long that it's actually
53:17
a huge pain to switch because of the
53:19
apps. Any chance Chris could share
53:22
how his transition has been going and
53:24
may he share how he set up simple apps such as
53:26
notes, calendar, and reminders. Well, I
53:28
would just refer you to our sister show, Linux Unplugged.
53:30
I think Chris has done a rather excellent
53:33
job over there of documenting his,
53:35
what is it, Graphene? Is that what you
53:38
losers call it over there? My
53:40
Graphene journey? Yes. If
53:42
I recall, it started in November
53:45
because I was at AWS re-invent in Vegas
53:48
back then. Still using it, still got it.
53:50
Yeah. Yeah, it has, you know what,
53:52
it's funny. You're gonna roll your eyes at this hard,
53:55
but you know what, can you guess what app I miss the most
53:57
out of all of Apple's apps?
53:59
Don't tell me it's the blue bubbles.
54:02
No, it's Notes. Really? OK.
54:04
I know. It's silly, but
54:06
in the last couple of iOS releases, Notes
54:09
is really, really competent.
54:11
I mean, you can share Notes. You can do
54:13
collaborative editing. You can take pictures
54:15
and store them in the Note. It supports Markdown
54:18
editing. It can capture text from inside
54:20
the Note. It has searching capabilities,
54:22
and you can search that text. It has tags. It has folders.
54:25
And you can export it all out to Markdown
54:28
using a third-party tool. It's really
54:31
everything I need, because I often just need,
54:33
like, the other day I was working on the car,
54:35
and I just wanted to get the label
54:37
off the battery.
54:38
I don't want that in my photo camera roll. I don't want
54:40
that backed up to the server. I don't want that on my slide
54:42
show system. I don't want that. You
54:44
know, I don't know what. I just want it for 10 minutes.
54:47
Or I want it in a year when I need to look up the battery
54:49
again.
54:50
And I want it in a Note. I want it in a
54:52
damn Note. And that's what Apple notes it. So I haven't
54:54
really solved that, but
54:56
Quillipad gets pretty close.
54:58
Q-U-I-L-I-PAD. It syncs
55:00
with Nextcloud Notes, and it gets
55:03
me really, really close. I like that. And
55:05
then Davix 5, you've got to
55:07
have that to sync with Nextcloud. You basically end up using Nextcloud
55:09
to do a lot of the iCloud stuff.
55:11
And that's how you solve it. What's a guy
55:13
got to do to get you drinking that Obsidian sauce,
55:15
huh? Oh, got it right here. I got it right here.
55:18
Oh, yeah. No, I'm using Obsidian. But Obsidian
55:20
isn't great for image-based stuff, you
55:22
know? Yeah, that's true. That's true. So I use it
55:24
for my actual oil change
55:26
notes and things like that. I put an Obsidian. But
55:29
it was great. It was the other day. Like, I did a bunch of work
55:31
on my car in the summer. And I thought to myself, when
55:33
did I last? Because I've got a track day coming
55:35
up in October. Thinking, when did I
55:37
last change the oil? Which
55:39
event was it before or after? And I couldn't
55:41
remember specifically how many track days
55:44
the oil in the engine has. And
55:46
I went and looked in my Obsidian, and sure
55:48
as you know, a few hundred miles ago. And
55:50
I changed it the day before. So it's done
55:52
one day on track, this current oil change. And I'm like,
55:55
oh, past Alex, I love you. Thank you
55:57
so much.
55:59
wife's the wife's sticker on her windshield that I put
56:02
on there when we changed the oil fell off. I'm
56:04
like, where's your sticker at? What sticker? Oh, yeah. Oil
56:06
change sticker. What are you talking about? The oil change
56:08
sticker I play. Where'd it go? I don't know.
56:10
So open up obsidian. There's the date.
56:12
There's the mileage. Okay, we're good.
56:14
Now just to finish Dimitri's question,
56:17
if you did want to catch up more about Linux unplugged
56:19
and Chris's journey with ditching
56:21
the big G in the sky,
56:24
it started at LUP 486. Goodbye, Google. Click
56:27
in the show notes.
56:28
Now we got some boosts this week and
56:31
our baller boost was spam proof at F
56:33
E A dot S T this week. He came in with 60,000 57 sats
56:36
using pod verse and he has
56:39
a hot tip because we've been getting what do
56:41
you guys use for personal finance management? He
56:43
says GNU cash. It's not great,
56:46
but I've been self hosting for over 10 years
56:48
now and it does a pretty good job of not breaking
56:50
with each new release. That's nice.
56:52
What does that, is that an indictment of the current state
56:55
of software or what? Yeah, really.
56:57
The headline feature is it doesn't break.
56:59
You know, there is also a real practical
57:02
matter of going with something
57:03
like GNU cash. It's not super flashy,
57:06
but it's been around for 226 years.
57:09
It just is steady as it goes and
57:12
it's open source and
57:14
it's established at this point.
57:15
So I got to, I got to give a plus
57:18
one to the GNU cash recommendation. Thanks,
57:20
band proof. And thanks for being our baller.
57:22
Leaky canoe came in with 50,000 sats using
57:24
the index. Hey gents, thanks for the great show.
57:27
I'm just getting started with home assistant.
57:29
What wisdom can you impart on a newcomer to
57:31
into this deep rabbit hole?
57:33
Also, what communication protocols do you choose
57:35
when you build out Z wave Zigbee?
57:37
Also you could add matter, wifi, thread,
57:41
uh, cues.
57:43
What wisdom can we impart on a newcomer?
57:46
Hmm. Start
57:49
small. It's very tempting
57:51
to order 800 devices and try and
57:53
do it all at once. True. You don't need to do
57:55
that. But as we talked about earlier in the show,
57:58
quite often you'll get halfway through doing.
57:59
doing a project, whether
58:02
that's an ESP based build,
58:03
or whether that is some kind of light
58:05
switch swap out and you'll figure out that, oh
58:08
actually the bulbs in the ceiling aren't compatible
58:10
with this type of dimmer switch and there's
58:12
gonna be roadblocks. So I would just take it slow,
58:15
do
58:16
maybe a room at a time or a certain
58:18
type of, do lighting all at
58:20
once or something like that, or
58:22
start off with climate or something really simple,
58:25
low hanging fruit where the stakes are pretty low.
58:28
I think I've talked about this in the past
58:30
where if you're doing something like home security right
58:32
off the bat and you're figuring out all the home
58:34
assistant nuances,
58:37
the stakes are high-ish in that
58:39
if you screw up, could potentially leave your house
58:41
unlocked overnight, which probably
58:43
you don't want to do that. Whereas if the lights
58:45
turn on at 2 a.m., okay, it's a bit annoying, but
58:48
nobody's gonna be actually
58:49
hurt or otherwise. Yeah,
58:53
hopefully. Yeah, I wonder what is
58:55
your thought on
58:56
doing a base of Buntu or
58:59
Sentos or whatever, Dix, and
59:01
running Home Assistant Core in a container
59:04
versus going with the whole Haas operating
59:06
system supervisor setup as for a
59:08
beginner?
59:09
The inbuilt app store is
59:12
super powerful. I don't use Node-RED
59:14
hardly at all anymore. I used to use it a lot when I
59:16
was in the beginning. So Node-RED
59:18
is a more visual-based way of writing
59:21
automations.
59:22
I got into that several years ago before
59:24
Home Assistant made their automation
59:26
UI a lot better. There's
59:28
still a place for Node-RED if you wanna
59:31
do some really complicated stuff because you can drop
59:33
to JavaScript if you're so inclined
59:36
as part of that workflow.
59:38
But I really like the VM
59:40
appliance aspect of it. I feel like
59:43
if I want to move my Proxmox
59:46
host, or I wanna do some maintenance on my main Proxmox
59:48
host, I can just snapshot that VM
59:51
and transfer that Q-COW file to
59:53
a different box and bring it up, no problem.
59:56
Whereas if it's on a physical piece of hardware and
59:58
it goes pop,
59:59
then... And if I'm not
1:00:01
in the house, it's
1:00:03
more tricky for me to recover from
1:00:05
that situation. The other thing I would say
1:00:07
is make sure you've got a proper backup. So I've been using
1:00:10
for the last several years the Google Drive
1:00:12
backup plugin. This takes a snapshot
1:00:15
inside the environment of
1:00:17
the VM
1:00:18
of the entire home assistant configuration, including
1:00:20
add-ons and all the rest of it. And it backs
1:00:23
up to Google Drive. You can configure how many snapshots
1:00:25
it keeps and it rotates
1:00:27
them out every seven days for you or whatever you want to do.
1:00:30
Yep. Also a plus one on that recommendation.
1:00:32
You can also go in there and have it do like, Hey, I'm about to go
1:00:34
do an upgrade. So do a backup for me
1:00:36
and immediately send that off to Google Drive. So
1:00:38
that way if anything goes wrong, I can bail out.
1:00:41
Also you asked about communication protocols,
1:00:43
Z-Wave, Zigbee, et cetera.
1:00:45
You know, man, that
1:00:47
is a hard question to answer. It kind
1:00:49
of depends, bro. Like depends on
1:00:51
your home
1:00:52
because I have both Zigbee and Z-Wave
1:00:54
and there's things I like about both of them. Zigbee
1:00:57
is an open standard. It is going
1:00:59
to also probably, you'll find cheaper devices
1:01:02
so you can save some money because
1:01:04
it doesn't require a certification and Zigbee
1:01:06
is kind of being folded into matters. So it probably has
1:01:09
a really long future. Z-Wave
1:01:11
is a proprietary standard you have to be certified
1:01:13
on, but that means that in order for devices
1:01:16
to get certified, they have to pass a certain
1:01:18
level of QA
1:01:19
and it's 900 megahertz versus 2.4
1:01:22
gigahertz for Zigbee. For me, 900 megahertz
1:01:25
just works better, goes farther, does more,
1:01:27
is more reliable. I tried to switch over
1:01:29
to Zigbee and went back to Z-Wave because
1:01:32
it's just 900 megahertz I think at the end of the day works better
1:01:34
for me.
1:01:34
I would try to put as few devices on Wi-Fi
1:01:37
as possible, absolutely fine to have
1:01:39
Wi-Fi devices. I've got plenty of
1:01:41
them. If you already own some, it's fine.
1:01:43
But ultimately, I like to
1:01:45
have everything on Z-Wave or Zigbee as
1:01:47
much as I can, especially things that are sensors, switches,
1:01:50
and that kind of stuff. And also,
1:01:52
if you can, when you're getting smart plugs, buy ones
1:01:54
that have energy monitoring built in from the beginning because
1:01:56
then you get all kinds of great data you can use later on.
1:01:59
ZP writes
1:02:01
in number one, I
1:02:03
own one domain and
1:02:05
now I just feel inadequate. You
1:02:09
know, it was priced me is how
1:02:11
many people boosted in just saying they own one domain.
1:02:14
Yeah. Yeah. Chris, you are not normal. My
1:02:16
friend. Well, how many domains? I mean, you have
1:02:18
a lot of domains, right? Uh, I
1:02:20
bought a couple this week actually. Yeah.
1:02:23
We actually tried to buy self hosted dot
1:02:25
forum this week.
1:02:26
Uh, but unfortunately that domain has taken
1:02:28
with, with a lemme instance right now and
1:02:31
the admins aren't interested in serving it. So
1:02:33
I would love it if we could do self hosted dot forum,
1:02:35
but alas, that's
1:02:37
why we just got to use, we just got to standardize
1:02:39
on dot lol. Yeah.
1:02:40
I had
1:02:42
to put that in there because first of all, it makes
1:02:44
it a great booster. And second of all,
1:02:46
one domain really dude, somebody
1:02:49
out there must have more than 10 or 15, right? You
1:02:51
got to admit it.
1:02:52
I, I feel, I feel like
1:02:55
a domain lush. I mean, it
1:02:57
might be, it might be 80 domains.
1:02:59
I don't know. I mean, it's not that many. It's
1:03:01
a lot though.
1:03:02
It's, it's enough that they were considered an asset
1:03:05
of the business when I sold the business. I'm
1:03:07
going by the American pie rules here. You
1:03:10
take the number, the girl
1:03:12
says and
1:03:13
times it by three and the boy says
1:03:16
and divide it by three. Lol
1:03:19
Salvatore came in with 10,000 cents. First
1:03:21
time booster, long time lover.
1:03:23
Tail scale question for the wizards. Is there
1:03:25
a way to have an
1:03:26
SSH only connection go over my
1:03:28
tail net?
1:03:29
It has my work machine locked down and sometimes
1:03:32
I want to connect to a guacamole server at home.
1:03:34
You might be looking for something like corkscrew.
1:03:37
This isn't a tail scale specific tip. Although
1:03:39
of course it will work over tail scale. I
1:03:41
don't know if your admins
1:03:43
permit VPN traffic. Some firewalls
1:03:46
are clever enough to detect that kind of stuff and
1:03:48
block it at the firewall level outgoing.
1:03:51
What I used to do when I worked for a bank, what's it
1:03:53
called when, when the thing expires
1:03:55
like Lance Armstrong, like seven years or whatever,
1:03:58
where
1:03:58
you admit to a crime long enough.
1:03:59
after that. Mason- Statue of limitations.
1:04:02
Mason- Yeah, I think that applies here. I think
1:04:04
I've not worked for that company for like six
1:04:06
years, seven years now. Mason- I think it's not statute,
1:04:08
it's statute of limitations. That's what it's statute.
1:04:10
Mason- Statute of limitations. Okay. Well,
1:04:13
when I worked for the bank in London, they had quite
1:04:15
a restrictive firewall policy, which of course being
1:04:17
a bank you would expect. And I was getting
1:04:20
into, you know, Linux in a big way and
1:04:22
SSHing around all over the place back then. So I
1:04:24
started running my SSH servers on port 443. Mason-
1:04:27
Clever boy. Mason- Because a lot of
1:04:29
encrypted
1:04:29
traffic goes over 443, so it becomes
1:04:32
just another encrypted stream. And
1:04:35
so if you use, there's a command,
1:04:37
which you can put in your SSH config file
1:04:39
to tunnel all of your
1:04:42
SSH traffic out over this tool called corkscrew.
1:04:45
And then it will go out over port 443
1:04:48
and look like just normal HTTPS traffic.
1:04:50
And that gets around quite a lot of sneaky
1:04:53
firewalls if you ever need that trick.
1:04:55
Mason-
1:04:55
Yeah, there's so many fun ways to play with SSH.
1:04:58
Let us know how it goes, Lulz Abertur. I think you're
1:05:00
going to have some fun.
1:05:02
VT52 and Faraday Fedora
1:05:04
boosted into say, pork bun.
1:05:07
They like pork bun as a DNS
1:05:09
registrar. They say they're headquartered in Portland, Oregon.
1:05:12
So for Chris, it's almost like buying a domain
1:05:15
from the next door neighbor. Chris- And you have to
1:05:17
have it checked for damp every few years as well, big
1:05:19
Portland, right? Mason- Portland,
1:05:21
potentially vandalism these days, but oh,
1:05:24
I kid.
1:05:25
There's a Seattle Portland rivalry in Seattle's better. Bobby
1:05:27
comes in with 6,000 sats. My
1:05:29
umber node broke. So Oak is down.
1:05:31
That's how he sends his automatic boost.
1:05:33
So I'm sending them by hand. Containers are still a bit of
1:05:35
a challenge for me to troubleshoot. I'm not really sure where
1:05:37
to go next with self hosting and Bitcoin
1:05:40
and Lightning. Do
1:05:41
I do something like Umbral, which is one of those you install
1:05:43
it and it has an app store and you install all the apps from
1:05:45
a container? Or is there maybe another
1:05:47
less black boxy way I should go?
1:05:57
or
1:06:00
some sort of desktop install and
1:06:02
play around with it on the command line and start
1:06:04
there run containers on the command line play around
1:06:06
with Docker compose because that's
1:06:09
what like these things are doing like start 9
1:06:11
and umbral and there's a bunch of others I was looking at
1:06:13
a couple others this week they're giving you really
1:06:15
nice front ends they're doing the app discovery
1:06:18
which is nice but when you click install
1:06:20
they're just kind of pulling down a Docker compose file
1:06:23
and then pulling down containers and firing
1:06:25
them up and when you understand how that works it
1:06:27
makes it pretty easy to troubleshoot any of these or at
1:06:29
least you can go in and look at how they
1:06:31
run and be like oh I'm not comfortable with this this is a mess
1:06:33
and you can bail
1:06:35
yeah I mean I got a bunch of tips over in perfectmedia
1:06:38
server comm which by the way
1:06:40
I should say in the last episode I put a call
1:06:42
out to the audience to say hey send me some
1:06:44
money please to support the website and
1:06:47
boy did you guys respond
1:06:49
I had nearly $400 in donations
1:06:51
come in in the last two that's awesome which was
1:06:53
amazing it's more more donations than I've ever
1:06:55
had for any of any of my projects previously
1:06:58
let's should cover some of the run costs of the server for a
1:07:00
while yeah it's about 15 months it's great I've
1:07:02
run of runway which is the first time perfectmedia
1:07:04
server comm has ever not just been out of my own pocket
1:07:06
so thank you so much to anybody that
1:07:09
donated and
1:07:10
I really really appreciate it so
1:07:13
on perfectmedia server comm there is a containers
1:07:15
section where I sort of walk through Docker and Docker compose
1:07:17
and all that kind of stuff you could also have a look
1:07:19
in my github repo which I'll put a link to in the show
1:07:22
notes for all the various containers that I
1:07:24
run
1:07:24
personally and I run that
1:07:27
through an Ansible configurator to spit out the Docker
1:07:29
compose file that may or may not be
1:07:31
too advanced for you in which case there
1:07:34
are tons of Docker compose examples
1:07:36
for those similar apps on the internet or
1:07:38
just join our discord and ask for some help and we'd be happy
1:07:41
to help.
1:07:41
Our last boost this week that makes it in before
1:07:44
we have to run is from Gene Bean 19,998 sats and he sent
1:07:46
me some pictures
1:07:50
of a traditional American
1:07:53
rotary phone where you put your finger in the thing
1:07:55
and you around
1:07:57
and he has done an integration with
1:07:59
his VoIP system and
1:08:01
he could pick it up and do voice
1:08:04
commands to Home Assistant on an old
1:08:06
classic rotary phone. He
1:08:08
sent me all the pictures in Matrix and it is so
1:08:10
neat and he
1:08:12
he pointed us to the VoIP integration
1:08:14
for Home Assistant
1:08:15
which is how he's kind of making all this happen. He has
1:08:17
a little bridge adapter
1:08:18
with a PoE adapter so the whole thing's
1:08:20
powered and it all just is just so great.
1:08:23
I remember I was probably a teenager
1:08:26
at this point but my
1:08:28
stepmom brought home a
1:08:30
rotary phone from she used to work
1:08:32
in the doctor's office and I think that they
1:08:34
were having a clean out or something and she brought home
1:08:36
this rotary phone and you know
1:08:38
I was a bit of a gadget
1:08:41
head even at age 14 or whatever
1:08:43
it was. And I remember she put this thing down on
1:08:45
the kitchen table and said to me how would you
1:08:47
how would you use this? How would you dial a number on this
1:08:49
phone? And I look at it and I'm poking the
1:08:52
buttons thinking I have
1:08:54
no idea. And I sort of got
1:08:56
the idea that the dial on the front sort
1:08:58
of moved and then I was like wait I
1:09:00
have to drag it all the way around
1:09:03
and then wait for it to go all the way back
1:09:05
and then do the next number the same way. Like
1:09:09
I want to dial 999 that's going to take
1:09:11
me a while. Yeah I
1:09:13
loved them though just playing with it. I
1:09:17
just like that I liked playing around with it. They had such
1:09:19
a wonderful like Land Rover
1:09:21
esque mechanical engineering
1:09:24
like clunk to them didn't they? Yes
1:09:26
they were they were very clunky technology.
1:09:29
Also I'm going to put a link in the show notes. Gene
1:09:31
Bean sent along a link for home cam
1:09:33
for a home kit and if your cameras are home kit
1:09:35
compatible
1:09:37
it gives you like a dashboard
1:09:39
of bringing all your camera feeds into a single
1:09:41
pane.
1:09:42
It looks really cool so I'll put a link to that in the show notes.
1:09:45
That's all the boost for this week because of
1:09:47
time but thank you everybody who boosted and we we
1:09:49
keep all of them in the boost part
1:09:51
and our doc and we share them with the whole team so everybody
1:09:53
sees all of them we had 19 boosters
1:09:55
and we pulled in 230,909 SATs. Thank
1:09:59
you everybody.
1:09:59
who did boost in, we read all of them. And if you'd like
1:10:02
to boost the show, you
1:10:03
can get a new podcast app, podcastapps.com.
1:10:05
Let's see pod friend is on there. Fountain,
1:10:08
Castamatic, podcast
1:10:10
index is on there. Podverse, lots
1:10:12
of different apps out there. Podcastapps.com or
1:10:15
keep your app, getalby.com, top it off
1:10:17
and then go to the podcast index and boost in and we'll read your
1:10:19
message
1:10:20
on a future show. And you're supporting us directly.
1:10:22
And
1:10:23
of course, thank you to our members,
1:10:25
our SREs that are making the show possible
1:10:27
and supporting the ongoing production. You get an ad free
1:10:29
feed with
1:10:30
bonus content, a post show.
1:10:32
That's it. Self-hosted.show slash SRE.
1:10:35
And don't forget about our upcoming meetups.
1:10:37
Linux Fest Northwest is of course happening
1:10:40
any day now. And there'll be
1:10:42
a bunch of last minute shenanigans. I'm sure.
1:10:45
Uh, we're thinking about doing some kind of a live recording
1:10:47
on October the 20th. We're not quite sure of
1:10:50
the details or whether we'll move it
1:10:52
to the Saturday whilst we're at the fest. Who
1:10:54
knows? Who knows? It's all up in
1:10:56
the air. We could do a live show at the fest.
1:10:58
That's not a bad idea, Alex. Yeah. I
1:11:01
don't know. Talk more about that.
1:11:02
We'll see. I mean, I don't know about you, but I kind of like
1:11:04
having an editor. Yeah, that's true. Well, you
1:11:07
can still edit. It's just going to be yikes.
1:11:10
Not, uh, not, it won't be his fault, but
1:11:12
it'll be noisy. Just be a little noisy.
1:11:14
Yeah. Yeah. So anyway, if you
1:11:16
want to find out more about that meetup.com
1:11:18
slash Jupyter broadcasting, as well as Linux
1:11:21
Fest Northwest.org, you can
1:11:23
go to self-hosted.show for all different places to
1:11:25
get in touch with us. And I have a
1:11:27
links site at alex.ktz.me. And
1:11:30
you can find me in the matrix. I'm at jupyterbroadcasting.com
1:11:33
slash matrix at Chris LAS. We got a self-hosted
1:11:36
room or two over there. We got a whole bunch of chat rooms. It's
1:11:38
really kind of a pop in place.
1:11:40
Come join us on the Federation. Thanks for
1:11:42
listening. That was self-hosted.show slash 105.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More