0:00

Once again today we're joined by Brent.

0:02

Hello Brent, how are you? Hey Alex, I'm good.

0:04

How are you guys? Well we're fantastic now

0:06

that you're here. It's always nice to have a little maple syrup flair. And

0:09

you know what? I just got back from El

0:11

Salvador and there was no maple syrup

0:13

to be had. Except for at Wendy's.

0:15

But you don't want their maple syrup. Makes sense. You

0:17

made it back in one piece, huh? I did. I did.

0:20

I'm feeling... Towards the end of the trip I was actually feeling pretty

0:22

good. I didn't really want to go. It's a

0:24

good school figure, right? But that's

0:27

just sort of how it goes sometimes.

0:29

But I'm glad to be back because we've got a packed

0:32

episode. Not only do we have a special guest

0:34

coming up, the core developer behind Image. But

0:37

Alex, you got some new gear

0:39

in the Homelab. It's been a veritable

0:42

circus of UPS drivers the last couple

0:44

of weeks. Not only did I get a new 3D

0:46

printer, I got one of the Bamboo Lab ones. Which

0:48

is awesome, by the way. I

0:51

also got my 45 Homelab

0:53

server review unit sat on the desk here

0:56

behind me.

0:58

Big and chunky boy on the desk.

1:00

It's like a 4U box. I'm

1:03

trying to

1:04

describe to you through the

1:06

medium of audio just how big this thing is. It's

1:09

like one of those fractal, defined

1:11

size cases. Just flipped on its side and designed

1:14

to go in a rack. They do ship

1:17

rubber feet to go with this thing if you want to stand

1:19

it upright vertically. One

1:21

of the things I really like about it is on the front

1:23

there's a pair of captive thumbscrews. To

1:26

get inside, they're on the front

1:28

of the case. Imagine it's in a rack situation.

1:31

You've got two captive thumbscrews that you twiddle.

1:35

Once they're loose, you just push the top of the case

1:37

back a little bit and pop it up. Then you've got access

1:39

to all of the internals of the case. They

1:41

sent me the fully loaded unit. The

1:43

one with the 6-core Xeon CPU in it, 32GB of RAM, 1TB NVMe

1:45

SSD.

1:50

Also, the 15 pre-wired

1:53

drive bays, which is just fantastic.

1:56

In this case, the build quality is the

1:58

first thing that struck me.

1:59

It is absolutely flawless.

2:02

This thing is built like a tank, really

2:04

thick. I assume steel, metal

2:07

of some description, I assume it's steel. Really

2:09

thick, good quality machining, no

2:11

tool marks. The powder coat job is fabulous.

2:15

It's up there with System 76's

2:17

build quality from when we took a tour of their factory

2:20

a few years ago. I don't know about you Chris, but I'm feeling

2:22

all sorts of envy. I know, I'm thinking,

2:24

boy, this sounds like a pretty nice rig.

2:26

I actually do think I'd wanna have it standing out

2:29

on the table or something. Maybe I wouldn't want it in the rack. Within

2:32

arm's reach, like Alex, you just give it a slap to

2:34

prove it's actually there. Yeah, so you can give it a good smack, yeah.

2:37

So that's the aesthetics of it. Have you taken

2:39

it much further than that? Just first impressions I assume

2:41

so far. Well, the CPU that's in there is

2:43

a Xeon 3204 from 2018 with

2:47

a 1.8 or 1.9, the internet can't quite agree on

2:52

the core base clock of this CPU. 32 gigs

2:54

of RAM that's in there is ECC memory

2:57

and the one-tier SSD is an

2:59

NVMe one, as I mentioned. I

3:01

find it really interesting that they went with a Xeon CPU

3:04

and there's obviously some reasons why you would

3:06

do that. If you look at the PCIe lanes

3:08

that Xeon affords you, it basically

3:11

enables them to put the HBA

3:13

card directly onto the Super Micro motherboard

3:16

that ships in this thing. It's an

3:18

X11 SBH motherboard. I opted for the one

3:20

with the SFP plus 10 gig pair

3:22

of NICs in it. They do make a

3:25

standard RJ45 variant

3:27

as well. Alex, I heard you say 2018 in there. Does

3:31

that seem quite old these days? Yes,

3:35

and there is no way around that. It's a six-core

3:37

CPU with no media

3:39

encoding engines either. So there's no quick sync with

3:41

this chip. That single

3:43

decision led me down a bit of a path

3:46

of thinking who is this box for? Because

3:49

when we had Doug and Mitch on a couple of weeks ago, talking

3:52

about how it's big, strong, fast, stable

3:54

storage, all that kind of stuff, the

3:56

performance for just storage

3:59

is really... But

4:02

the minute I have a server on in my house

4:04

24-7, I'm looking at what else can

4:06

it be doing. I want to be doing image,

4:08

I want to be running image on it and doing a bunch of machine learning

4:11

tasks on it. I want to be doing Jellyfin

4:13

or Plex or whatever and doing a bunch of transcoding on it

4:15

maybe for remote

4:18

media streaming, that kind of stuff. And

4:20

I find the choice of a 6-core

4:22

CPU with a sub 2GHz clock

4:25

speed from 5 years ago peculiar.

4:29

I understand that the choice of motherboard

4:31

is probably what precipitated the choice of

4:33

CPU because

4:36

certain motherboards have certain features like HBA's

4:38

built into them like this one does and it supports

4:40

up to 2TB of RAM because it's a Xeon

4:43

chip. All this kind of stuff.

4:45

There's a lot of very valid reasons for

4:47

doing a Xeon. However,

4:51

if you're putting this in a home lab, it's in

4:54

the name. It's probably going in my home.

4:57

And for most people at home, you

4:59

are using a box with 15 hard

5:01

drive slots to store some kind of media,

5:04

probably. You might be a video

5:06

editor. I can't imagine that's a huge

5:09

use case, although certainly there will be

5:11

a bunch of YouTubers that use this as

5:13

their primary video editing

5:15

storage system. But I've got to imagine

5:17

that most people building these or using these

5:20

cases to build servers

5:22

are doing things like Unraid and TrueNAS

5:25

and Perfect Media Service, dare I say

5:27

it, all that kind of stuff. And so why wouldn't

5:29

you put a CPU in it that

5:32

has hardware media encoding?

5:35

It really confuses me as

5:37

to the identity of the product

5:39

a little bit. I bet it is predicated

5:42

on the motherboard. And I wonder if

5:44

just describing which motherboard and CPU

5:47

are coming in the system sort of solves this problem so then

5:49

you just know before you pull the trigger here. And

5:52

maybe some folks don't care about QuickSync. I

5:54

mean that is a thing, right? You can just buy the

5:57

chassis on its own, the case on its own, for...

6:00

about $800. So the fully loaded unit that

6:02

they sent me, which full disclosure I did

6:04

not pay for, the full unit

6:07

is $2,000. They sell a version of just the case

6:09

for about 900 and then the

6:12

case and the power supply for 900 and then

6:14

just the case is about 800. Right.

6:17

So you could, I guess, get that one and put your own motherboard

6:20

and CPU in. Absolutely, yeah. And if you look at what Wendell's

6:22

up to on F41TEX, that's exactly what he's

6:24

doing. He's putting some like AMD Epic

6:26

server chip in there and balls

6:29

to the wall faster than VME it can

6:31

possibly be. It's kind of nuts. I

6:33

want that one. I want that one. So

6:35

this got me thinking about, you know,

6:38

for $800, let's review just

6:40

the chassis on its own for a second.

6:42

For $800, what else could I buy?

6:45

And the obvious choice

6:47

for me at least is the Rosewill LSV4500U,

6:51

which is a bit deeper and a bit longer and it

6:55

only has 12 drive bays

6:57

with the hot swap loading out the front

6:59

rather than out the top like this one. So I

7:01

mean, that's kind of a wash. Like if you want to load

7:03

the hard drives in and out the front of the case without having

7:06

to pull it in and out on rails, maybe

7:08

you consider that a pro. If you like

7:10

that kind of thing, then the 45 drives with 15 versus 12 is a pro for

7:15

that one. So that's really a wash

7:17

for me. The 12 versus 15 slots. I

7:20

don't know. I think I'd probably rather have the 15 drives

7:23

and then I don't also have to put them in sleds. Yeah,

7:25

that is the thing. I mean, the drive

7:28

bays at the front of this thing are tool-less. So

7:31

that's pretty cool. But you know, the other

7:33

thing I think about is five years ago when this CPU

7:35

was birthed into the world, hard

7:38

drive sizes were 8, 10, 12-ish

7:40

terabytes. And so 15 slots

7:44

back then was, air quotes,

7:47

only 150 terabytes, which

7:49

is still a hell of a lot of terabytes. Now, 20

7:53

terabyte hard drives are $300

7:55

a pop, which, okay, is not cheap, but

7:58

in terms of a rational purchasing decision,

8:00

you know, you could quite easily put 300 terabytes

8:03

in this box now in that same

8:05

density, which is if

8:08

you're using 300 terabytes at home,

8:10

I'd love to know what for

8:12

if it's a totally legal application.

8:16

That's all I'll say. Oh, I could see

8:18

it. I could see it. Yeah. These

8:21

days, especially with large models and, you know,

8:23

somebody wanted to bring other media for a media

8:25

production company. I could see it. I

8:28

mean, again, you're talking this is a device that lives

8:30

between the consumer like

8:33

the NAS for consumers and

8:35

the enterprise, right? This is

8:37

a 15 drive bay system. And

8:39

in the case with the fully loaded one, it comes with a 2018 processor. But

8:42

that kind of makes sense if you're coming from an enterprise design

8:45

standpoint. So it's for very serious folks.

8:48

And an $800 starting entry point

8:50

for just a chassis and the backplane for

8:53

people that are pretty serious seems

8:55

about right when you can put 15 drives in there and you

8:58

can put your own motherboard in there if you want and your own CPU.

9:01

Yeah, I mean, that's that's kind of nice.

9:03

It is a bit expensive, but it sounds really nice for

9:05

people that are serious. It is absolutely

9:08

a really nice bit of kit. It's

9:10

a premium bill quality. And

9:13

I've got no complaints about it. Really. I

9:15

just I just wish that it was five

9:18

or six hundred dollars, not eight hundred.

9:20

Oh, yeah. That would be a totally different equation

9:23

for me because at that price

9:25

point, it becomes the de facto like

9:27

standard unit of deployment for home

9:30

labbers and folks looking to build on raid

9:32

servers and and that kind of thing. They came

9:34

out with the HL-12. You know,

9:37

the HL-12, 12 bays, 500 bucks,

9:39

because I don't really know if it's the slots you're after.

9:42

It's you want something that's quality, right? You're going

9:44

to put this in your home. Craftsmanship. You're going to

9:46

have your home pictures on there. Right. Craftsmanship. I

9:49

think that counts for something. It does, which

9:51

leads me to a little bit of a

9:53

software experience out of the box that I had some issues

9:55

with too, unfortunately. So the image

9:57

they shipped to me had Rocky Linux. the

10:00

SSD and the

10:02

first thing I did was try to log in as

10:04

you can imagine. So first of all I had

10:06

to look for the IPMI password

10:09

which was printed on a tiny little sticker

10:11

on the motherboard. Their documentation unfortunately

10:14

told me it would be printed on a white sticker on the side

10:16

of the case that wasn't present so I had

10:18

to go spelunking on the internet and find

10:20

that Super Micro Motherboards as

10:22

of 2019 put this unique password

10:25

on the motherboard. And I spent

10:27

a while trying to get that

10:30

password in with the correct username of

10:32

admin and then realising that admin had

10:34

to be all caps. It

10:37

was case sensitive so it literally took me half

10:39

an hour to turn it on. That's on me,

10:41

some of it, you know, but some of it's on the documentation

10:44

too. Once I then got the system

10:46

booted up I was presented with

10:48

a GNOME login screen and I was like,

10:50

ooh username 45 drives. So I look in the documentation

10:53

for the HomeLab server which

10:56

is like 100 pages long. I had

10:58

to get to page 61 before I

11:01

found out what that username and password was

11:04

and it was wrong. It told me the username

11:06

was root and the password was correct

11:09

but it was for the user 45 drives

11:11

not the user root so you know some teething

11:13

issues there and nothing major,

11:16

just some small niggles to fix in the documentation.

11:19

Then I loaded up the terminal once

11:21

I got logged in and thought cool right time to update

11:23

packages as you do like a reflex after

11:26

you've used Linux for a long time. Absolutely.

11:28

And I had to do dash dash skip broken

11:31

dash dash fix something like

11:33

I had to pass three different parameters

11:36

to ignore all sorts of shenanigans that was going on

11:38

with DNF in order to fix a

11:40

bunch of out of the box broken packages which was a

11:42

bit disappointing. Hmm that

11:45

is. I'm not you

11:47

know I'm not sure if I'm actually even just thrilled

11:49

with the choice of Rocky. I do again I can see

11:52

it from the enterprise perspective. Well I think

11:54

they ship Rocky because they can do cockpit

11:56

modules and they spent a bunch of time

11:58

on their Houston UI. which is actually

12:01

based on top of cockpit and they've shipped

12:03

some custom modules in there which let

12:05

you see which drive is in which physical slot

12:08

on the system Some

12:10

proper system integrator level stuff, which

12:12

is brilliant. That's exactly what I want to see at this

12:14

price point And then the

12:17

only final Niggle I had

12:19

was I loaded up Firefox because

12:22

it has a desktop for some reason this server comes pre-installed

12:24

with a desktop Whatever, that's fine

12:27

And Firefox had someone else's browser

12:30

history in it Presumably the tech that was setting

12:32

up the server like it was looking at super

12:34

micro fan control and a bunch of other stuff that

12:36

They googled and gone through the github link and downloaded

12:39

the script and I was just thinking

12:41

to myself That

12:43

cheapens it a little bit because I would have expected

12:45

them to have a pre-made Golden

12:48

image with all of that stuff removed and you know,

12:51

I come from the enterprise world with red hat for a long

12:53

time Where building those kinds of experiences

12:56

out of the box experiences is exactly

12:58

what tools like satellite and ansible have been

13:00

doing for? Years and years and years

13:02

and I just I wonder why that wasn't

13:04

done here. It was a custom bespoke

13:07

edition Just for me. I Would

13:11

imagine too like that image is probably gonna be something

13:13

they're developing quite a bit. They'll be probably

13:15

updates their shipping Yeah, but knowing

13:17

you I don't I don't really see you wanting to run rocky in

13:19

production Are you planning to keep rocky install

13:22

on there? Are you gonna try going with something else? I

13:24

want to try putting nix on this bad boy There

13:26

you go. I actually want to do a full-on nix

13:29

OS server One of the guys

13:31

I work with has been showing

13:33

me some really ninja level stuff

13:35

with nix And system d n

13:37

spawn containers and how he's configuring

13:39

that with a like a helper script as part of his

13:42

flake setup I'll get in some details

13:44

into that over the next few weeks months because

13:46

it's it's pretty complicated and I'm still learning

13:48

it myself but it looks amazing

13:50

and Isolated containers

13:52

on the host as part of a nix declarative config

13:55

yada yada yada. Yeah, it's

13:57

pretty cool stuff I'm feeling it. That's what that's why

13:59

I want want to do our server here at the studio as

14:01

Nick's. We almost went Proxmox and then I started

14:04

thinking about it and like, wait a minute, we've missed

14:06

the window Chris, it's mid November. We should have done Nick's

14:08

November. I know,

14:11

I know. Nick's our new year maybe. But

14:14

so, I mean, it really should have no problem, right? It's

14:16

a standard PC, it's an

14:18

X86 box. Yeah. It should just load right

14:20

on there. It's just a Xeon box with a

14:23

CPU and memory. It's

14:26

just a Linux box, right, at the end of the day. I

14:28

know it's early but I mark my

14:31

words, in a couple of years, hardware

14:33

vendors like 45 Drives

14:36

and 45 Homelab and I also

14:38

think iX Systems, they

14:40

will realize that

14:42

to achieve the level

14:44

of quality and reliability they want,

14:47

it will be in their best interest to

14:49

ship this hardware with Nick's OS. I had

14:51

an absolutely fascinating conversation

14:54

with an individual in El Salvador who's shipping hardware

14:56

to stores and by a

14:58

process of elimination, they started with Debian

15:01

and went through everything else and by a process of elimination,

15:03

they're now using Nick's OS because of its quick

15:06

ability to deploy, recover and do all that stuff

15:08

that they need. I think a lot of hardware vendors

15:10

are going to get there. So, you know,

15:13

you'll have to keep us updated on what it's like to run it on this machine.

15:15

Well, I absolutely will. So, I think the way I'm going to split

15:18

my Homelab up now is I'm going to keep

15:20

my existing server with the i5-8500

15:23

CPU in it and Quick Sync. I'm

15:25

going to keep that for all the mundane sort

15:27

of media serving tasks that I have in the house

15:30

and then this one is going to be straight

15:33

up like business suit and tie, you

15:35

know, photo storage, next

15:37

cloud, that kind of stuff that, you

15:40

know, I'm going to throw a few hard drives in there

15:42

and do ZFS, all that kind of stuff. So, I think

15:46

for me it'd be nice to have that separation of duties

15:48

as well between like a serious server and

15:50

a fun server. Do I need it? No,

15:53

but it is fun as hell. Alex,

15:55

I know often when you're reviewing these

15:58

kind of units, energy... consumption

16:00

comes to mind for you. Have you measured this yet? Now

16:03

whilst I don't have 15 hard drives to load

16:05

this thing up with what I can tell you

16:07

is what the Standby power consumption

16:10

is because this thing has a BMC built into it for

16:12

proper IPMI because it's a super microsurfer

16:14

motherboard So just turned

16:17

off plugged in with just the BMC

16:19

on is 23 watts Which

16:21

if you consider that is already more than most

16:24

of those small form factor like 1 liter

16:26

PCs on its own True

16:28

gives you an idea of things but

16:30

better than those Dells that

16:33

we have plugged in Yeah,

16:35

you can't throw two terabytes of RAM

16:37

into a micro PC Can you I mean the more we talk

16:39

about it? I really see how this fits between like a one liter

16:42

PC home lab machine and also

16:44

like a huge used like super micro

16:47

Or old Dell or HP rack unit just

16:49

powered at rocky Linux just at the desktop

16:52

No hard drives installed nothing like that

16:55

roughly roughly 120 watts It's

16:58

a Xeon at the end of the day and that has Some

17:02

ramifications with with energy usage. So if you

17:04

compare that to like the i5 8500 for example, that's drawing 10 to 15

17:09

watts for idle. Okay, you've got a bunch of hard drives

17:12

in that box too But that's again,

17:14

we're be really nice to have click sync in that CPU

17:16

so you can throw a few more tasks out Yeah,

17:19

now it's a server. So they've put server grade

17:21

fans in it The CPU itself

17:24

doesn't have any fans on the heatsink So

17:26

that should give you an idea of the level of noise that

17:28

you can expect from the other fans in the

17:30

chassis, too You could legitimately

17:33

condition the air in your house with

17:36

this thing If you put a HEPA filter on the front

17:38

of it, the airflow is quite impressive So a

17:40

big thanks to 45 drives for sending that unit

17:42

over for review and you'll be hearing more about that in the

17:44

upcoming episodes That's

17:46

a great way to support the show you get a hundred

17:49

devices and you can see that tail scale truly is simple and secure

18:00

but we love that it's built on WireGuard. It'll

18:02

get up and running in just minutes. TailScale's

18:04

so great when you're traveling. I

18:07

mean, it's fantastic if you just work from home and

18:09

you wanna have a secure connection

18:11

to maybe a VPS and something that's running on your local

18:13

machine and a phone. You can kind

18:15

of create your own mesh network across all those

18:18

different networks. That's really handy. But

18:20

man, oh man, when I'm traveling,

18:22

oh boy is it great. Knowing

18:25

that every photo I was taking, every contact

18:27

I was saving, everything was being synced securely

18:30

over TailScale to my next cloud box, that's

18:33

a level of peace of mind that I just haven't

18:35

really been able to describe before when I'm joining like an

18:38

airport Wi-Fi in El Salvador. I

18:40

just love knowing that all my information's

18:43

protected by TailScale. And TailScale

18:45

Traverse Double NAT like it's nothing.

18:48

And when I'm switching between networks, going between

18:50

weird cellular providers I've never heard of and

18:52

onto an Airbnb Wi-Fi, TailScale

18:56

never missed a beat. My connections

18:58

between my servers, my computers, my mobile

19:00

devices, even some of my

19:02

appliances was just bulletproof. I

19:04

was managing my power input on my

19:07

Victron system using a subnet

19:09

router on my LAN and my RV connected

19:11

over cellular. And I'm doing

19:14

it all from my phone while I'm traveling.

19:17

It really is a great resource too for moving data

19:19

back and forth. They have TailScale Send which made it really

19:21

easy to copy files back and forth. If you wanna

19:23

share resources with folks, you can either integrate

19:25

it with your enterprise ACLs or

19:27

if you're just an individual, some friends, they got a system that

19:30

works for you. It'll completely

19:32

work when you're separated by firewalls and subnets.

19:35

It's just, it really is the chef's

19:37

kiss of networking. It's where I wanted to

19:39

see WireGuard go when I first heard it get described

19:41

way back in the day when it was just a twinkle

19:43

in the colonel's eye. From arch servers

19:46

to mobile devices, you've gotta try it out, deploy

19:48

it, and see how fantastic it is to build your

19:50

own private mesh network. Just

19:52

go to tailscale.com slash self-hosted.

19:55

You can try it for free for up to 100 devices, really

19:57

make sure it works for you, and it's a great way to

19:59

support the show. So it's tailscale.com

20:02

slash

20:02

self hosted

20:05

Well two weeks ago. We talked about the verge cast

20:07

and their big self hosting Podcast

20:09

episode and in that they interviewed

20:12

Alex who is the lead developer for the image

20:14

project now image is Ostensibly

20:17

a Google Photos self hosted

20:19

replacement or at least that's what it's trying to be So

20:21

Chris and I sat down with him to talk through

20:24

the future of the project and the roadmap and also,

20:26

you know Open source sustainability and that

20:28

kind of thing So we

20:30

finally got Alex on the show the developer

20:32

of image and it's kind of great timing

20:35

because version 1.8 6 just

20:37

came out and Alex welcome

20:39

to self hosted and sorry we haven't had you on the

20:41

show before now. No worry Thank you for

20:44

having me is the pleasure to be here to talk

20:46

to you guys and to talk about image Well,

20:48

the new release came out just a couple of days ago as

20:50

we record. It's got some nice features

20:53

I think specifically like around sharing right that

20:55

is correct Yeah, so

20:57

this new release we add the unified

21:01

Sharing for partners. So

21:03

now if you have multiple partners, you

21:05

can enable that option so all

21:07

of the access of the partners always is

21:09

also showing on your timeline and This

21:12

make browsing Experience

21:15

a lot better so you don't have to switch to

21:17

the sharing tab and then click on each partner

21:20

to see the new Media

21:22

but now just in your timeline But

21:24

there are a lot of caveat as this is the

21:26

basic version. The searching is

21:29

not including the partner media

21:31

yet

21:32

and

21:33

The only thing that share on that

21:36

media is the the location

21:39

For now, that's fantastic. Because if you

21:41

think about how most people actually use a photo

21:44

solution, right? My wife is

21:46

taking pictures of my kid I'm taking

21:48

pictures of my kid and we probably both want

21:50

to see pictures of our kid, right?

21:53

And we say it's a family unit thing and

21:55

a lot of the self hosted solutions

21:57

to date don't really take

21:59

that much multi-user within a

22:01

single household, a single family into

22:04

account. So I'm absolutely thrilled to hear that you guys

22:06

are. Thank you. And I think this

22:08

is coming from a lot of my

22:11

own use case as well, because scratch

22:13

the own, scratch your own itch. That's the best way to

22:16

develop software, in my opinion. Yeah. And then we

22:18

add the new map tire server to moving

22:23

away from using the free tire

22:25

server from OpenStreetMap as

22:27

the usage policy. I

22:30

thought that was really funny when I looked at it, because

22:33

you were like, well, when we were just a small scrappy

22:35

project, we could get away with using it for

22:37

free. But now we're kind of a big deal. So

22:40

we kind of need to do this properly by the book.

22:42

Yeah. The community, there

22:44

are some concerns that

22:47

asking us that we should move away as soon

22:49

as possible to not

22:51

overuse their free

22:53

resources. And we take

22:56

this change and we move away to a

22:58

different map tire server and

23:02

it's also giving us the opportunity to build

23:04

our own layers. So to

23:06

make it look slicker on the

23:09

dark and the light mode.

23:11

I think it does. I think it does.

23:14

So I want to talk about the audacity

23:16

of the project for just a moment, because the idea

23:18

of trying to replace Google Photos kind of seems like

23:21

an impossible task. And yet version

23:23

one dot eight six getting pretty

23:25

far now with face recognition, sharing

23:28

support, the location.

23:30

When I was looking at my pictures from my trip, the

23:32

new map stuff is really slick. So you

23:35

are kind of in a transition phase now to

23:37

a big project, right?

23:39

Yeah, I would say so. Just

23:42

to look at this core

23:44

member,

23:45

because they kind of group the

23:47

new user coming in by

23:49

day. So when you look at it, you can

23:52

see it grow. And

23:54

when I first started this core, we might

23:57

have one or two new people

23:59

come coming in. the server every

24:01

day and now when you look at it, it

24:03

can reach like 20-40 people. And

24:07

then the star

24:09

on GitHub is also the indication

24:12

of how the project is getting more popular.

24:15

And also, more

24:18

often than I would admit that I would browse

24:20

the internet to see how people are talking

24:22

about image. And yeah, we've

24:25

seen a lot of chatting and

24:28

most of the girls are coming from a worth

24:31

of miles. Yeah, it's coming up online all

24:33

the time now though. I feel like I see it getting mentioned

24:35

every week at least by some different group online

24:38

saying they're using it to replace Google Photos. People

24:41

are recognizing it for a genuine

24:44

replacement. It's pretty exciting. I

24:46

just wanted to say congratulations.

24:47

Thank you. My goal is

24:50

to build something for myself and

24:53

find image maybe what

24:56

it is. You

24:58

and I have crossed paths a couple of times on Reddit

25:00

I think in the past and talked about your funding model

25:02

in particular in terms of a

25:05

sustainable model for an open source project.

25:07

I mean, we've seen Home Assistant with Nebuchadse for

25:09

example build a company around

25:12

an open source project.

25:14

I think

25:15

we've also discussed in cross paths that

25:17

something like this photo

25:20

solution is so big

25:22

in scope that you do almost need a company

25:24

to support it. You've obviously

25:26

done amazing work and you've got some fantastic contributors

25:29

to date but I wanted to ask you what the plans

25:32

are for the future of image as a

25:34

project and also as a sustainable project

25:36

in open source. So I think there are

25:39

one thing that I

25:41

want to not never pivot

25:44

off is to not making it into

25:46

some type of paywall like

25:49

some of the open source projects

25:51

starting out as open source and then they got funding

25:53

from outside

25:55

investor and then they got that money pressure

25:58

to eventually have to make money. of

26:00

pivot of the first original idea.

26:04

So I don't want

26:06

to do that for image at all. I want to make it as

26:08

something good, something people

26:11

enjoy using. So that is something

26:13

that we would never like to move away

26:15

from. The car team

26:17

has talked about maybe eventually after

26:20

the service, the app is more stable.

26:23

We could think of building a

26:27

hosting solution for other people. So

26:29

people that would like to use image but don't

26:31

want to host it themselves. So

26:35

I think that is the, this might

26:37

be the only way right now to

26:40

make it as a sustainable

26:43

business. But by not

26:45

making it as the main

26:48

income source at home, for

26:51

all of us, I think that's making

26:53

it as a sustainable project

26:55

that keeps growing. Because we don't

26:57

have to depend on it for our livelihood.

27:00

You still have your day job and you still

27:03

dedicate your time outside

27:05

of job in order to

27:07

build image. And it's a

27:11

joy to do that as something,

27:14

I think the whole core team really enjoy

27:16

because we have such a nice group to

27:19

chat about technology, code,

27:22

and just how to

27:25

beat Google. You

27:27

love to hear that, right? And

27:29

really, it's great that while you grow

27:32

the project and get it to that point where it's stable

27:34

enough that it could be a product, that you do have the

27:36

support and income of a day job. And

27:38

maybe one day those two things merge and

27:41

image becomes a bigger part of your income source.

27:43

Yeah, and that's what I

27:46

try to achieve is when

27:48

it gets to that point where I can

27:50

replace the income source of image

27:52

into my own, that

27:55

I can support my family and the growth of

27:57

my family, then that would be fantastic.

28:00

work on something you love, you make good

28:02

things for people, and then you also get to earn money

28:04

from it. Yes, that would be absolutely fantastic.

28:08

How could people support the project today?

28:11

Today, there is only donation

28:14

means to support the project besides

28:16

testing and reporting bugs to

28:19

help us develop the project

28:21

better. You can support the project

28:23

financially from GitHub sponsor

28:26

or from Leapro

28:29

Bay Pay or buy

28:32

me a coffee. I drink a lot of

28:34

coffee, so keep the coffee

28:36

coming. In

28:39

the Vergecast interview that we talked

28:41

about on the last Self-hosted episode, you're

28:44

now a super famous Big Deal developer

28:47

and the big time on the Vergecast podcast.

28:49

I wouldn't say so. You

28:52

talked a little bit about your architectural

28:54

decisions about splitting it out into microservices

28:57

and a bunch of Docker containers and stuff like that. Is

29:00

there anything you wish you'd done differently now

29:03

or are there any big changes coming down the

29:05

pike? Yes, so the goal is

29:07

to

29:09

consolidate some of the services

29:11

like the proxy container and

29:14

the web container to put that into the

29:16

server container to make it less

29:18

scary when people are looking at the Docker Compose

29:21

file because those can be done. But

29:25

when I started it, it makes sense

29:27

to split it up at that moment.

29:30

And now if we might have found a way

29:33

just today to make the

29:35

web not server-side

29:37

rendering in anymore while still

29:40

keeping that really snappiness when

29:43

using the app.

29:44

So with microservice architectures,

29:46

you have the potential for remote

29:49

workers and stuff like that. Correct. I

29:51

mean, I would love to see it where I could distribute the facial

29:54

recognition. So I just dumped, for

29:56

example, like three or four terabytes worth of photos

29:59

into image. in the last couple of months because

30:01

you guys are awesome and added external library

30:03

support, which is amazing. Mwah, chef's

30:06

kiss. Thank you. But it

30:08

took like two or three weeks for it

30:10

to chew through all of those images

30:12

that I had. I've got lots of CPUs

30:14

in this house. It would have been great if I had an easy

30:16

way to distribute that task throughout

30:19

all my systems. Yeah. So

30:21

actually the idea of

30:23

breaking down microservices,

30:26

different containers, handle different things

30:28

is exactly to serve this use case. So

30:31

as of now, you can run

30:33

the machine learning on

30:35

completely different machine. And if

30:37

you put a frontend of load balancing

30:39

on it, then you can distribute it on

30:41

multiple nodes because we are

30:44

not accessing the file

30:46

on that container, but you

30:48

are sending the file

30:51

to those containers and then it would do

30:53

the referencing and then sending back the data

30:55

that it needs. The microservice

30:58

is a little bit more involved

31:01

because you would need to be

31:03

able to see the file system similar

31:06

to the server so that it can access and

31:08

do the transcoding stuff. But if you

31:11

have network mounted, you can also do

31:13

that already. Pretty typical distributed

31:15

network storage replication

31:17

problem there. Yeah.

31:20

But I think the

31:22

architecture has been designed

31:25

exactly for that

31:26

purpose. And we have a lot of people

31:28

that already do that. I think that will help you

31:30

guys as you move towards eventually

31:33

a hosted service. You will be able to pick the

31:36

cheapest VPS provider that week. Yeah. Do

31:39

you know it maybe? Yeah, we got a great

31:41

coupon code you can use. Well

31:44

I think for me and probably a lot of people

31:46

listening, I know Alex too, this is one of the tools

31:49

that has really been

31:51

the key to de-googling so

31:53

much of my private information. And

31:56

so privacy and security is super

31:58

important. I'm just curious to know your thoughts. around

32:01

enabling sharing but keeping people's

32:04

images safe and secure and how that kind of

32:06

works from a back-end perspective. Is your

32:08

question more about at

32:10

home? Yeah, like say if I wanted to share an image

32:12

with somebody, would there

32:15

maybe be like a back-end service like down

32:17

the road, some sort of intermediary that would be

32:20

the go-between to share the image so I wouldn't have to expose

32:22

say like my home machine or something? I'm

32:24

trying to think of ways of sharing images where I don't

32:27

I don't link someone to my image instance. That's

32:30

so challenging at the moment

32:32

but anything might

32:35

have a solution. We're

32:36

on video chat with Alex, dear listeners and

32:38

I can see the cogs turning right

32:41

now live as you're thinking what

32:43

if we had some kind of like remote cloud

32:46

you know. Yeah, that's what I'm wondering. VPS

32:48

like linked up or something you know that's

32:51

great idea Chris. Feature requests right

32:53

there. Turkey Days coming in and I'm thinking something

32:56

that I could like a page that would be generated

32:58

that I could share with family but I wouldn't have to link

33:00

them

33:01

they wouldn't all be necessarily loading it from my direct server.

33:04

I see so in that page you would talk

33:06

to the image server at home but that is

33:09

abstracted in the background and then

33:11

it would use the image API

33:13

in order to pull those just like a proxy

33:16

right just like a proxy pay. Eventually

33:18

this is something that we want is to build a plugin

33:21

system for the

33:23

people to build things around

33:25

an image. The API

33:28

is compatible with a lot

33:30

of the languages so we used open

33:33

API so it can generate

33:36

SDK for Python for JavaScript

33:39

for Rust if you want

33:41

for anything that you could think of right and

33:44

then they can use that in order to talk

33:46

to their server the image instant

33:49

and now one of the we have one really

33:51

good tool I would like to mention here is called image

33:54

go is from Simulat

33:56

it has this court hand

33:58

handle

33:59

This is a CLI

34:02

tool. You point it to your Google

34:05

Tech Al. You don't even have to extract

34:07

all the zip file that you download. It

34:10

unzip it and then it pars all

34:12

the metadata from that JSON file

34:14

that Google gave you. And

34:17

then upload everything into image

34:20

and create the album that comes

34:22

in your Google account. So it's just like one

34:25

click go. That's

34:27

so neat.

34:28

Yes. I love that. See, that

34:31

plugin system, if that really extends, it could be

34:33

a solution because people could build things around that. Because

34:35

when I really A-B, when I miss from Google Photos,

34:37

it's like a way to share photos and I'm not

34:40

like sharing any of my personal infra. But

34:42

something down the road, like a plugin

34:44

system, there's all kinds of ways to solve that. I'm

34:48

thrilled to hear that you guys are thinking more about that too. It's

34:50

going to be an exciting future as an image

34:52

user. Absolutely is. Now,

34:54

I have another question for you, Alex. Ever

34:57

since the beginning of the project, there's

34:59

been a banner in your documentation

35:01

and on your website that says, here be dragons.

35:03

Do not do anything serious

35:06

with this project because it's still under very active

35:08

development. Do you have any senses

35:10

or time scale as to how long before we can

35:12

actually start to air quotes

35:15

depend

35:16

on image? You know, in this court, we

35:18

have a funny way of saying that this is the any

35:21

question is a

35:23

prohibit question. We don't

35:25

talk about that here. Yeah,

35:28

fair enough. Don't commit to deadlines on air

35:30

in particular. I get it.

35:32

We put that because part

35:34

of it, this image is very active in

35:36

development, right? It's like just this

35:39

last release, the first time ever

35:41

we pushed out the first

35:43

release that if you don't, if you're not on

35:46

the same version, you cannot access

35:48

the instant from your phone at

35:50

all

35:51

because we changed the underground,

35:54

the underlying API. That's

35:57

the only way to resolve the difference is

35:59

to lock out. and lock back in on the same version.

36:02

Things like this could happen because we

36:04

are striving to go to the

36:07

industry standard for the

36:09

code base. If anything that we see

36:11

can be improved for the better future

36:13

of image, we will do

36:16

that. And since there are

36:18

many things that we still can improve,

36:21

we don't want to prohibit us to

36:23

backward compatibility

36:26

at the moment. But people

36:28

are striving to get to the point of feature

36:33

freeze or when the core

36:35

architecture is more

36:38

stable, we

36:41

would like to eventually

36:43

remove that

36:44

banner.

36:45

I think we all just see the potential of the

36:47

product and want to start using it. I

36:49

say product. It's not a product. Project.

36:51

And we just want to start using it today because

36:54

I'm real fed up of the

36:56

cloud nonsense and giving our machine

36:59

learning, overlord, model, so

37:02

much data. So I think that's

37:04

the only pressure really from us is that we kind

37:06

of wish it already existed because the pressures

37:09

that forced you to create it in the first place

37:11

are quite powerful. Yeah, very tempting.

37:14

They're tempting fruits, aren't they? Alex,

37:16

I'm going to keep using it even with that banner on there.

37:18

So there. And thank you so

37:20

much for creating it. And thank you to the entire team and everybody

37:23

that works on it and contributes. Absolutely. I

37:25

know a lot of us in the self-hosted community are just

37:27

extremely grateful. Yeah, I would like to shout out to the team

37:29

too. Everybody in the team

37:32

is great. They come from different backgrounds.

37:34

We have students from Europe. We have people

37:37

working in the application security, some

37:40

11x developer.

37:42

So the whole team is great.

37:44

And people are very

37:47

welcoming and willing to teach

37:49

each other. So I want

37:52

to keep this mindset growing in order

37:54

to attract more contributors

37:57

and just to make it a good place to be.

37:59

not just from using the application, but from developing

38:02

the application as well. And so if you'd like to contribute,

38:05

you can go to github.com and image-app,

38:07

I believe, is the place to go. You've

38:09

got a couple of hundred contributors already. A few more

38:11

wouldn't hurt, right? Yep.

38:15

So a big thank you for me and Chris for joining us on

38:17

the show today. And I'm sure this won't be the last

38:19

time that you check in with us. But until

38:22

next time,

38:22

thank you very much. Thank you guys. Thanks, Alex. Thanks,

38:25

Chris. You guys have a wonderful night.

38:28

So, I'm Chris

38:30

H. Linode is where we host everything in

38:32

the cloud. And now Linode is part of

38:34

Akamai, the Akamai. But with all the

38:37

tooling, the UI that we

38:39

like, the CLI that's super handy, all

38:41

that stuff that you can use to build, deploy, and

38:43

scale in the cloud, that's

38:45

still there. It's just now they're combined

38:48

with the power of Akamai's global reach.

38:50

They are the premier network out there. And

38:53

Akamai is expanding

38:55

their cloud services, expanding

38:57

tooling, but keeping things affordable,

39:00

reliable, and scalable

39:01

for users of any size, an enterprise

39:04

or an individual project. And something we've taken

39:06

advantage of in the last month is the expansion of the data

39:09

centers. They're expanding worldwide to give us more

39:11

access to more resources so you can grow

39:13

your project or your business. So

39:15

why wait? Go see why Linode is the only cloud

39:17

provider we recommend. Experience the power

39:20

of Linode, now Akamai. Go

39:22

to linode.com slash SSH. That supports

39:24

the show, and it gives you a $100 credit

39:26

so you can really kick the tires. So

39:28

go see how Linode, now Akamai, can help scale

39:30

your applications from the cloud to the edge.

39:32

That's linode.com slash SSH.

39:37

Well, at that time of year, Brent, you messaged me saying,

39:39

I want to buy some hard drives, and I couldn't

39:41

refuse the opportunity to sit down with you and

39:43

discuss hard drives. It's literally

39:46

my specialist favorite subject in the whole world.

39:49

Once a year, you get all giddy about getting new

39:51

hard drives, and I thought I'd join in this year. So here's

39:54

part one. Brent and I sat down and recorded

39:56

a few different segments to use over

39:59

the holiday season. and discussing

40:01

his journey into buying hard drives

40:03

and purchasing considerations, all that kind of

40:05

stuff. I'm going to put part two

40:07

in today's post show, which I believe, Chris,

40:10

we have a special offer for our members. It

40:13

is the Black Friday season after all. So

40:15

if you've been thinking about becoming a member and support the show and

40:17

you want to get that post show, use the promo

40:19

code Black Friday and you'll get $2 off a month for

40:23

a whole year of the SRE membership.

40:26

Just go to selfhosted.show slash SRE

40:28

and it's promo code Black Friday, all

40:30

one word. And then you get the post show and you get a

40:32

post show every single episode. So

40:36

all the way back in 2019, October 2019, four years ago, Brent,

40:39

you and I sat down with

40:44

Antonio, the developer for Merger

40:46

FS, and you professed to me in that

40:48

episode and public record,

40:51

no less, that you wanted to take

40:53

storage more seriously. And here we are four years

40:55

later and I think it's

40:57

about time we cashed that check. It's

41:00

long overdue, Alex. And I admit it

41:03

had been on my mind for four years previous to that

41:05

as well. So I am

41:08

in bad shape because we were talking about this before

41:10

we pressed record. And you said

41:12

to me, I've been buying drives with ZFS in mind

41:14

in pairs for like the last eight

41:17

years. No more than that.

41:20

Since my photography career required

41:23

it, which is, I had

41:25

been doing photography for 18, so I've got

41:28

paired one terabyte

41:30

drives and paired 500 gig

41:32

backup drives. Basically anytime

41:35

I'd buy drives, I would try to pair them because

41:37

I knew one day I would be smart enough to actually

41:39

do something about it.

41:41

And that day is today, is it?

41:43

I'm hoping it is, but my

41:45

level of success so far is challenging

41:48

at best. Well, okay. Let's

41:51

start at the beginning. You mentioned photography.

41:53

Now, I know that's a career path that was

41:56

huge for you previously, but now obviously

41:58

with your next cloud work. I would imagine

42:00

your data generation is a little,

42:03

you know, the velocity is a little slower on that side

42:05

of things. What kinds of stuff

42:07

are you looking to store at the moment?

42:10

Yeah, it's a good question. I think

42:13

most of my data, at least the

42:15

big stuff at this point,

42:17

is archival, to be honest. Like

42:19

it can just sit there and

42:22

won't get access very often, then it's a read-only

42:24

probably kind of scenario if ever.

42:26

Okay, so that leads me into my next question. If

42:29

it's archival, do you

42:31

genuinely need it to be always

42:33

on? Always accessible.

42:36

Hmm. I hadn't even considered that because

42:38

I had just pictured all of my storage being

42:40

in one box. But tell me

42:42

what you have in mind. If it's an archival situation,

42:45

it's not beyond the realms of possibility

42:47

that you've got a client ringing you up that you had

42:50

on the books 10 years ago that says, you know that shoot

42:52

you did in our mine, could you

42:54

find that particular, you know, shot

42:57

of Fred or whatever? I've

43:00

gotten those emails before. Yeah, I'm sure you have. And you think,

43:02

right, well if I just, you know, write down

43:04

what's on each hard drive, external hard drive

43:06

in a drawer, for example, that

43:08

could be a legitimate filing system for

43:10

that data. The issue

43:13

with that is Bitrot is a thing

43:16

and that keeping those drives just sat in a

43:18

hard drive, you know, they're not going to last forever.

43:22

So you could explore something like

43:24

Amazon's Glacier service,

43:26

which is a cloud based long term archival

43:29

situation. It's fairly cheap.

43:32

They do kind of hit you on the retrieval fees.

43:34

But if this is an archival situation only,

43:38

maybe it's a decent option. I hadn't

43:40

considered that at all as my primary

43:42

source of storage. I had only considered

43:45

cloud solutions as kind of a offsite

43:47

copy. I think mostly

43:50

because I have, I don't know

43:52

if I

43:53

trust them with the primary data.

43:56

Is that one way to look at it? But you're suggesting maybe

43:58

I have drives in a drawer and also... so

44:00

some data there, is that what you're suggesting? It

44:03

could be. I mean, the other option is you

44:05

just have all of that data running all the time

44:07

in your house, and it's always available on

44:09

top of ZFS so that you've got all the checksumming

44:12

and beautiful magic that ZFS gives you for data

44:14

integrity. And then you back

44:16

that up to Glacier. And the purpose for suggesting

44:19

that is it would reduce the

44:21

complexity for you somewhat in terms of having

44:23

to have a box at a remote house with, as

44:26

I found out fairly recently, having

44:29

a remote box in a different country or a

44:31

different side of the country even, even

44:33

if it's in a family house, it's

44:36

hard sometimes to get access to that. And the

44:40

economics of something like rsync.net

44:42

don't really work out. Seven

44:44

or eight terabytes in rsync.net is 70

44:47

or $80 a month. Oh,

44:49

I see what you're saying. Which pays

44:52

for a remote backup server in no

44:54

time at all. But Glacier

44:57

is, from memory, it's 10 bucks

45:00

for a similar thing or less

45:02

per month. Like I say, the

45:04

way they get you is on the retrieval. So if you try and do

45:06

an expedited retrieval, they'll charge

45:09

you $10 for a thousand API requests versus

45:13

just a standard rate of 3 cents per 1000 requests. Hence

45:18

the name Glacier. Because it's like,

45:21

only if the world is going wrong do you want to melt that thing.

45:24

Yeah, they've never actually published what the underlying

45:26

storage technology is. I suspect it's tape,

45:29

but we don't actually know. And so it would

45:31

make sense because I would imagine they have

45:33

those tapes in some kind of robot

45:36

automation system that can actually move

45:39

a physical piece of media in and out of

45:41

a specific location. And

45:43

if you're doing an expedited recovery, that probably

45:45

means some S hit the fan somewhere. And

45:48

you need that stuff back pretty quick. All

45:51

right, so let's leave Glacier for a second

45:53

and talk a little bit more about your local

45:55

hardware.

45:56

Desires, you sent

45:58

me a list of all the hard drives. you've got and let me just

46:01

let me just read this to the audience because it's

46:03

it's quite glorious I think you're going to enjoy this

46:05

is it really yeah I love I liked

46:07

it there is a Western Digital 8 terabyte

46:10

drive brackets shucked a

46:13

Western Digital red 4 terabyte drive

46:15

a Western Digital caviar black 2 terabyte

46:18

drive 3 seagate 1.5

46:21

terabyte drives 2 Western

46:23

Digital caviar 1 terabyte drives

46:26

and 2 500 gig drives

46:28

how many is that one two three four five

46:30

six seven eight nine ten ten drives

46:33

Brent that's quite a lot

46:35

unfortunately this doesn't include any of

46:37

the two and a half inch like external drives

46:39

that I've been carrying on all my travels to

46:42

do you know local backups as I go so

46:44

there are a couple more of those maybe we should throw

46:46

in the mix but they might not be useful in the setup

46:48

I don't know and then the next line in

46:51

in our little chat exchange was

46:52

but

46:53

the data duplication inefficiencies

46:56

galore etc so really

46:59

I have no idea quite how much storage I

47:01

have so if we add all

47:03

of that storage up dear listeners it comes

47:05

to approximately 21 and a half

47:07

terabytes but your guesstimation

47:11

on the use of that is somewhere

47:13

between 8 and 13 terabytes

47:16

is that right

47:17

yeah I think part of the challenge

47:19

here is that these

47:22

drives were never all available at the same

47:24

time and so they ended up being

47:27

generations of storage

47:30

for me you know someone fill up and then

47:32

they would go on let's call it long-term

47:34

storage and be replaced and

47:37

as they filled up depending on how

47:40

busy I was at the time or what it would just kind

47:42

of get slotted in and so the data is not

47:46

extremely well organized you know it's organized as

47:48

well as I could muster along

47:50

the way but because of that drive

47:52

swapping

47:53

there's also duplicates of stuff especially

47:56

with system backups and things like that there

47:58

are multiple multiple copies of that

48:00

kind of data and photography

48:02

projects. Sometimes on one set of

48:04

drives there's a project that was half finished

48:07

and on the next set of drives there's the whole project that

48:09

was finished or something like that. I

48:11

think if we approached

48:13

this intelligently and got rid of

48:16

all those duplicates, I would imagine there's way

48:18

less data than I think there is. But

48:20

to be honest, I'm just human and I have

48:22

no idea. You know what's interesting, I

48:24

think we've all been there. We've all found an old

48:26

hard drive in a drawer and gone, hmm, you know

48:28

what? I should back this up and then a couple

48:30

of years, elapse, and then

48:32

you find that same hard drive in a drawer and you go, you

48:35

know what? I should back this hard drive up. And

48:37

before you know it, you have that hard drive

48:39

in triplicate somewhere named slightly

48:41

differently, you know, every time. Well,

48:44

the thing for me is like,

48:46

this is just another iteration of my trying

48:49

to get that great solution. I mean, you've got

48:51

perfect media in the server, right? And

48:54

I've had various versions of

48:56

tons of hard drives in a box

48:59

and various versions of having it be

49:01

functional for me, but it never quite

49:03

got to the point where I felt really

49:06

comfortable about the data and that

49:08

it was, you know, being checked for parity and

49:10

that if a drive failed, that it wouldn't lose anything,

49:13

etc., etc. So I

49:15

still, despite my,

49:16

you know, knowing better and my efforts, I have

49:18

a long way to go, I feel like, which

49:21

is embarrassing, to be honest. Well, in

49:23

that case, my suggestion

49:25

is keep it simple.

49:28

Do a pair of disks

49:30

in a ZFS mirror and

49:32

then never think about it again. I mean, obviously monitor

49:35

them and all that kind of stuff. What I mean in terms of

49:37

you worried about like check summing

49:39

and all that kind of stuff and parity and then data

49:42

integrity, ZFS gives you all

49:44

of that out of the box because it doesn't

49:47

actually care about specific files.

49:50

ZFS cares about the blocks, the data

49:52

blocks, and it checks that if a specific

49:55

block is read from disk and doesn't

49:57

match the check sum it expects between the two mirrors.

50:00

it will go away and figure out what the right one is. And

50:02

then if it can't give

50:04

you the fully integral

50:06

data, it just

50:09

won't give you that piece of data at all. Which

50:11

sounds like a bug, but actually if you

50:14

ask me, it's a feature because it means that

50:16

you're never gonna get garbage out of a ZFS system.

50:20

The other nice thing about doing a pair of drives

50:23

like that is your unit of expansion

50:25

becomes really simple. And these

50:28

days you can buy 20 terabyte

50:30

hard drives for $300 each, which

50:32

isn't cheap granted, but

50:36

also for 20 terabytes of mirrored

50:39

storage, effectively 40 terabytes

50:41

of raw, 20 terabytes of usable.

50:44

That's actually not a bad deal at all. The

50:46

other thing about doing a pair is slot

50:49

density matters. You

50:52

might have, let's say four slots

50:54

in a server. So

50:57

let's put Brent in the shoes of five

50:59

years in the future Brent and this 20 terabyte mirror's

51:01

full. You think to yourself, well,

51:03

what can I do here? Well, you could just go and

51:05

buy another pair of drives of

51:08

whatever the in-vogue size is

51:10

that year. Let's assume it's 40 terabytes

51:12

in five years time. And

51:14

then suddenly you've more than doubled

51:17

your capacity, but all the old data is still

51:19

there in its old format for

51:22

you to move around between as

51:24

you see fit. The mental model,

51:26

it keeps it really simple to understand.

51:29

You know, that leads me straight into the

51:31

whole crux of the question that I have

51:34

today,

51:35

which feels like a bit of emergency. We're at

51:37

Black Friday season and a

51:40

bunch of drives are on sale and if you look at

51:42

them in Canadian dollars, I

51:44

have a potential to save like $200 per drive or

51:47

something with some of these sales. So

51:49

my question was, you know, you're suggesting

51:52

these, get the massive most drive

51:54

that I could possibly muster in my budget

51:56

and throw those as a pair

51:58

into the storage.

51:59

track that I have here. But it

52:02

gets me thinking about all these drives that I already have.

52:04

I've got an 8 terabyte, for instance. Can

52:07

I just buy an 8 terabyte drive and combine

52:09

it with that?

52:11

It seems like maybe

52:13

that would be far cheaper, and maybe that would

52:15

get me far enough where I could wait a year or two

52:17

and then get that massive pair

52:19

later. What do you think about the now versus

52:22

later question? The same logic

52:24

applies. Yeah. Get the biggest

52:26

hard drive your budget will

52:28

permit to pair with the old one.

52:31

Even if all you buy today is one 20

52:33

terabyte hard drive, for example,

52:36

pair that with the 8 that you've already got,

52:38

the 8 terabyte. Your mirror would

52:41

be an 8 and a 20. You'd be

52:43

wasting 12 terabytes today, which sounds

52:45

a bit stupid. But in

52:47

six months' time, when you come across another good

52:49

deal, and this just allows you

52:51

to do dollar cost averaging of hard drives,

52:53

I guess. You're buying

52:55

a hard drive now to make

52:58

the mirror and then in six

53:00

months' time, all you do is swap out the 8 terabyte

53:03

for the new 20, for example,

53:05

re-silver it, and suddenly then you've got the full 20

53:08

available to you. I had never

53:11

even considered wasting 12 terabytes

53:14

just for a couple months. It's a bit of a first

53:16

world opulent problem, isn't it? But

53:18

when you put it that way, when you're trying to

53:20

balance budget and maybe drive slots

53:23

and things like that, it

53:24

actually, in a strange way,

53:26

makes a lot of sense. But

53:30

it doesn't

53:31

make it easy to make a choice, does

53:34

it? No. The biggest knock

53:36

against ZFS really, versus something

53:39

like MergerFS, which you know I'm a great proponent

53:41

of,

53:42

is that

53:44

with ZFS you have to start with empty drives.

53:47

My question to you would be this 8 terabyte

53:50

drive, do you have somewhere to put that 8

53:53

terabyte drive whilst you build

53:55

the new array, the new

53:57

Z-pool?

53:59

It's a very... good point. I

54:01

will admit this 8 terabyte I also

54:03

may have

54:04

broke the connector on it so it's a little... Yeah,

54:06

it's easily done. So here's

54:10

the other thing right, we were talking about the data

54:12

you wanted to store and it's all kind of suit-and-tie

54:14

business-y data right? It's it's real

54:17

data that that cannot be replaced

54:19

stuff.

54:20

Yes.

54:22

Why not use your existing fleet

54:24

of mismatched hard drives to store

54:27

some slightly more replaceable stuff instead

54:29

using something like merger FS. Don't worry

54:32

about the parity on that and then have two

54:34

different tiers of data stores within the

54:36

same box using merger FS

54:38

and and that for one

54:40

side for the media and then ZFS for the suit-and-tie

54:43

data on the other.

54:45

I also hadn't considered that I'm

54:47

glad you're here with me. Yeah

54:50

and I like the idea of the peace

54:52

of mind of new drives because these drives you

54:55

know have

54:56

been around and just been sitting there as well

54:58

and they've traveled across the country in the

55:00

back of a trailer etc so my

55:03

confidence in these older drives

55:06

is also surprisingly low so the

55:09

idea of buying new drives feels I

55:11

don't know emotionally like a good thing but I

55:15

guess that brings me to another question of you

55:17

know if you go to somewhere like diskprices.com

55:20

to see what the best deal might be they just

55:22

by default show you a bunch of used drives

55:25

and those seem to have great prices

55:27

and they seem like enterprise drives and I'm just

55:29

curious your opinion of you

55:32

know if you shouldn't care about drives they

55:34

should just be replaceable then is that

55:37

a reasonable avenue to go down?

55:39

If you're running a data center yes. If

55:42

you are a man with less

55:46

than a dozen hard drives in his house

55:49

then each one actually matters at least

55:52

a little bit. You could argue that an off-site backup

55:55

makes that argument a little bit less relevant

55:58

but for every time a drive fails, it's

56:00

extra cost, it's extra cognitive load, it's

56:03

extra time watching and re-silvering, and

56:06

all that kind of stuff. So from my

56:08

perspective, you're

56:10

going to be saving probably 30% or 40% with

56:12

a used drive versus a brand

56:14

new one. But you just don't

56:17

know its history. There's

56:19

just no way to know how many times Linus Tech Tips

56:21

has dropped that particular drive, is there?

56:24

Oh, local drives. Yeah,

56:27

I guess you're right. The same applies. If

56:29

I'm not comfortable with the history of the drives that I have in

56:31

my own care, then I certainly shouldn't

56:33

be comfortable with the drives

56:35

that I know not the history.

56:37

There are certain things in a computer that I will absolutely

56:40

advocate buying used.

56:42

RAM,

56:42

definitely, because you can check that really easily. Really?

56:45

Yeah. CPUs, they almost

56:47

have only ever had one go bad

56:49

on me. Motherboards, again,

56:52

they either typically work or they don't, as long as you check

56:54

the CPU pins very carefully. Power

56:56

supplies, typically, they're

56:59

pretty reliable, too, as long as they've got all the original

57:01

modular cables with them. You don't swap them

57:03

out with a different manufacturer, because the pinouts can be

57:05

wrong, and then wrong voltages fry

57:08

things. Obviously, a case, as

57:10

well. Obviously, that doesn't really matter if

57:12

that's used or not. But hard drives

57:14

are the one thing, because they are so

57:17

delicate, and the tolerances are so small, and

57:19

they're so fragile.

57:21

I just don't mess about with them. Well,

57:24

that's great advice. Thank you, Alex. It feels like

57:26

now I just got to open my wallet, see what I get

57:29

shipped here this week.

57:32

45homelab.com, the HL15

57:35

is here. It's available for purchase. You

57:37

can get it in a bare bones thing, or you can get

57:39

it built up with a chassis and a backplane

57:41

and the PSU, ready to go, fully tested.

57:44

It's really up to you. Big, strong, fast storage

57:46

for the Homelab. Taking the enterprise

57:49

mindset, the lessons learned that 45

57:51

Drives has taken from all their years in the enterprise

57:53

and collaborating with the open source community, and

57:56

now building something for the Homelabber.

57:58

It comes with Rocky Linux. you'll have the ability to

58:00

install applications on it. Things like Image will

58:03

run fantastic on the HL-15.

58:06

And from what I hear, Alex has one arriving

58:08

soon, or maybe already arrived, actually. I

58:11

don't want to say too much, but chances are there's

58:13

going to be a review on the show. I'm really excited

58:15

because now we have something that I think

58:17

meets the bar, and it takes this

58:19

as seriously as we all do. So go try

58:22

it out. I think this is the perfect unit for anybody that

58:24

wants something robust that can grow with them.

58:26

It's the HL-15, and you can find it over at 45homelab.com.

58:31

It's here. It's based on your feedback, and

58:33

it's beautiful, big, strong, and fast at 45homelab.com. Well,

58:38

Alex, thanks so much for your insights. As always,

58:41

I

58:42

should just listen to what you say and just go and do

58:44

it. So thank you for that. Oh,

58:46

stop. But I bet you we've got a whole bunch of

58:48

booths we should go through.

58:50

We have a nice batch this week. We can't always fit them

58:52

all for runtime.

58:54

But Bamham 182 comes

58:56

in with 34,567 SATs using

59:01

Podverse this week. And they write, I

59:03

appreciated hearing about the backups as I'm building

59:05

a backup server. I've gotten an Elite

59:07

Desk 800 G3 SFF for $40 on eBay. That's

59:13

amazing. That's what I keep saying

59:15

about these one-liter PCs, man. They

59:17

are bargain of the century. And

59:20

it has two 3.5 MVME drives in it. I

59:23

bought the Easy Stores, which will

59:25

replace some drives in my main zPool. These

59:28

will go into the Elite Desk. If all goes as

59:30

planned, the box will go to a friend's house and

59:32

is off all the time. But it'll turn on at night

59:34

and sync the data through WireGuard and then

59:37

turn back off. I switched from PFSense

59:39

to OpenSense with all of the WireGuard

59:41

drama, and then from OpenSense to

59:43

NixOS because I became more and more

59:45

familiar with NixOS. Absolutely

59:48

no regrets so far. I love having full control

59:50

and not having things I don't need running 24-7. I

59:53

recently wanted to swap out a Dell Wyze I

59:55

was using for a T740. I

59:58

was able to get the T740 running. with a

1:00:00

very similar config, swap the boxes in it

1:00:02

without the family even realizing I had

1:00:04

done it. Nice job, B.

1:00:07

Nice. I have a question here.

1:00:10

Given these little one liter PCs and they just

1:00:12

sip power, why would you bother

1:00:14

turning them on and off at night like that? It's

1:00:16

probably not your power bill. It's not your power, right?

1:00:18

So even a little bit, still something, right? Yeah,

1:00:21

but I just worry about

1:00:24

mechanical hard drives, spinning up, spinning down.

1:00:26

Is that a consideration here? Or am I just

1:00:29

worrying for no reason? It is a consideration.

1:00:31

It adds significant complexity

1:00:33

to all of your monitoring as well. Also

1:00:36

the exit conditions of the backup scripts

1:00:39

itself, it becomes a consideration

1:00:41

to you. If the server's on all

1:00:43

the time, you don't have to have any of those considerations.

1:00:46

You just can assume it's always on.

1:00:49

And if ZFS tries to do a replication, it will

1:00:52

say, well, you're already doing a replication dummy. Don't

1:00:55

try and do another one. It's in the middle of one and it shuts

1:00:57

the computer down. And then

1:01:00

let's say you just took a trip, hypothetical

1:01:02

example, and you're dumping 100 gigs worth

1:01:04

of pictures onto your local

1:01:07

system. That could easily take

1:01:09

a couple of days or more to upload to the remote

1:01:11

site. And in that case,

1:01:14

you've got to plan for that one-off

1:01:16

eventuality, that unusual eventuality. But

1:01:18

you've still got to plan for it. And like I say,

1:01:21

it just adds complexity to the whole situation.

1:01:24

And reading the rest of this comment, you're

1:01:26

clearly a gentleman that appreciates the finer

1:01:29

things in life by ditching any

1:01:32

non-declarity of OS. So

1:01:35

look at us go. I

1:01:39

love it. That's right, the gentleman's OS. I

1:01:42

think it's a pretty solid setup. I totally get

1:01:45

turning it on and off to save power,

1:01:47

although I have a similar setup here. I have

1:01:49

my main workstation

1:01:51

on a

1:01:52

Zigbee Smart Plug. When

1:01:55

I arrive, I have an automation that turns

1:01:58

on the Smart Plug and my BIOS is set. to

1:02:00

restore power when the power

1:02:02

state comes back. And that's worked

1:02:04

for six months. And then when I went on my trip

1:02:06

and came back, the smart plug turned on. And

1:02:09

now every other time, the PC doesn't restore

1:02:11

power state. Just for some reason, still

1:02:13

set in the BIOS, I haven't changed a thing,

1:02:16

and it just doesn't restore power state. So that kind of thing

1:02:18

can happen too.

1:02:19

Now, Alex, this gets me thinking about, you

1:02:22

sent me a photo a couple days ago of

1:02:25

something you 3D printed for your one liter PC,

1:02:27

which to me on the surface seems ridiculous.

1:02:30

Oh, yeah. But

1:02:32

can you maybe give us more details?

1:02:35

I 3D printed a rack mount, a

1:02:37

one new rack mount for these little systems. Chris,

1:02:39

I'm going to just see your listeners. Here we go. Oh,

1:02:43

my goodness. This is amazing. So I've

1:02:45

got one of these little Dell boxes, one of these little

1:02:47

PCs in a rack mount

1:02:49

system now.

1:02:51

That is so cool.

1:02:54

And then obviously, a 3D printer can't fit 19

1:02:56

inches, which is a standard rack unit width. So

1:02:59

you split it in half, put some screws in the middle,

1:03:01

and suddenly you've got a 19 inch wide

1:03:04

one unit for two of these things. You

1:03:06

put two of them in there, one, yeah, and one. Oh, my gosh.

1:03:09

There's quite a few if you look on like principles or Thingiverse

1:03:11

or something like that. There's quite a few people that sort

1:03:13

of stack them on top of each other like this, and you can

1:03:16

fit an eight port switch in there

1:03:18

as well. So you can actually literally have like a five or

1:03:20

six half width U

1:03:23

rack, like a mini rack almost,

1:03:25

like a desktop home lab rack, just

1:03:27

3D printing. I need that for the ODROID.

1:03:30

I have yet to find a case I am fully satisfied

1:03:33

with, with the ODROID. And I just need a rack

1:03:35

unit that the top is exposed. That's

1:03:38

really great. Just buy a bamboo lab printer, dude, and be done with

1:03:40

it and get it over with. All

1:03:42

right. Johnny Castaway comes in with 26,235 stats using podverse. Johnny

1:03:47

Castaway? Isn't that great? And that's a great name.

1:03:50

He's a member. That brings me back. That

1:03:53

brings me back. Longtime listener. And

1:03:56

he says thank you for contributing to everybody who makes

1:03:58

my eight hour weekly commute. We

1:04:01

are the we are the commuters friend. He

1:04:03

says I was once at peace with arch then I

1:04:05

tried Nick's OS gave up

1:04:07

many times returned back to my old arch slippers,

1:04:10

but Nick's kept taunting

1:04:12

me like a beast Nick's has been tamed

1:04:14

now, and I think it's here to stay He says

1:04:16

my next endeavor is Nick's OS on the steam

1:04:18

deck any thoughts No,

1:04:21

do you want to mess up a good thing? Oh, it's a

1:04:23

UK. He also comes in from the UK Yeah,

1:04:26

all of this learning people must do these

1:04:28

days, right? Yeah,

1:04:31

it's funny because I was such a curmudgeon

1:04:33

when it came to Nick's as well and Chris brought

1:04:35

it up in the show and I

1:04:37

believe privately said to me Alex if you just listen

1:04:40

sometimes I identified Nick's years

1:04:42

ago Why couldn't we've done this in self? I

1:04:44

was like yeah, but I just didn't see it and

1:04:46

it's just one of those things like nobody

1:04:48

nobody can tell you how

1:04:51

good Nick's is nobody can Can

1:04:53

tell you how it's gonna change your

1:04:55

life You just have to experience

1:04:58

in it just has to click For you

1:05:00

and it's like that moment as a teenager

1:05:02

where your parents tell you you just don't understand

1:05:05

You're like well one day eventually

1:05:07

with enough experience you

1:05:09

realize just how powerful

1:05:12

that stuff can be You

1:05:14

know in Linux unplugged sometimes we get a hard time because

1:05:16

we're mentioning Nick's OS Far

1:05:18

more this year than well we ever

1:05:21

had But

1:05:22

it's just

1:05:23

because it solves so many issues

1:05:26

in a beautiful way I mean, there's that learning

1:05:28

curve, of course, but I

1:05:30

have heard just this week alone

1:05:32

like five to six different

1:05:35

people say Oh, yeah.

1:05:37

Thanks guys. You made me try Nick's OS now. I can't

1:05:39

use anything else and So

1:05:41

if you haven't tried it yet, I I'm

1:05:44

running it on this laptop in front of me and I was a curmudgeon

1:05:46

too It took me a while to come to it and it just like it's

1:05:48

just kind of nice It's kind of nice

1:05:51

another tool in the toolbox. I mean you look at

1:05:53

the the wider ecosystem

1:05:55

I mean the whole flakes adoption situation

1:05:58

is frankly. It's just a mess But

1:06:00

you've got companies like Determinate Systems with

1:06:03

Flake Hub coming along, pushing the agenda, and

1:06:05

taking the project by the scruff of the neck and saying,

1:06:08

Come on boys, sort this out. That's

1:06:10

exactly what it needs. And for me,

1:06:12

those signs are the green

1:06:14

shoots that I needed to know that the project was

1:06:16

going to be worth investing in in the long term.

1:06:20

Yep, I completely agree. Although I'll take the other side

1:06:22

just to be fun and say, Don't try it because it'll wreck you. Stick

1:06:24

with your distro. Stick. Don't change it. You'll

1:06:26

never be the same. The Bitcoin

1:06:29

Dad Pod comes in with 24,444 sats using Podverse. It

1:06:33

writes, Backblaze Cloud

1:06:35

is another remote endpoint for backups. I use Restick

1:06:38

and their S3 encrypted object storage. My

1:06:40

last bill was $4 for just two

1:06:42

terabytes of storage. Whoa.

1:06:45

That's attractive. Mm-hmm.

1:06:46

Noted. It says I've deployed Tailscale

1:06:49

at limited scale and I find it amazing. That

1:06:51

said, I'm concerned about ReLion on it

1:06:53

for as a third party now that it has

1:06:55

visibility into my network. They also seem like

1:06:57

an obvious point of regulation in a world of governments

1:07:00

obsessed with surveilling private online activity. Has

1:07:02

anyone ever tried to migrate a Tailscale set up to Headscale?

1:07:06

What are your thoughts on the potential risks? Well,

1:07:08

I can't speak to the migration to Headscale because

1:07:10

I completely trust the Tailscale

1:07:13

infrastructure. And here's why. If

1:07:15

you know anything about public key

1:07:17

infrastructure, you will know that that is

1:07:19

what WireGuard is based on top of. So the

1:07:22

idea that you provide a key, the remote server

1:07:25

authenticates based on that kind of stuff, and

1:07:27

it's very difficult to break that

1:07:29

level of encryption. Any

1:07:32

data that transits any Tailscale

1:07:34

asset is encrypted using

1:07:37

that methodology. But

1:07:39

the thing about Tailscale that kind of makes it magic,

1:07:42

and okay, yes, I work for them. Corporate

1:07:44

shill moment, hat on, is

1:07:47

most of the connections happen through the magic

1:07:49

sock. So they become peer to peer connections.

1:07:52

So very little data actually traverses

1:07:55

any Tailscale owned infrastructure. Any

1:07:58

point you can look at Tailscale, scale status,

1:08:01

look at all the nodes in your network, you can do tail

1:08:04

scale net check, which prints an analysis

1:08:07

of the local network conditions,

1:08:09

like all the local DERP servers, all that kind of stuff.

1:08:12

So when you are connecting two tail

1:08:14

scale nodes together, they initially

1:08:17

connect and report back through

1:08:19

the tail scale control server via

1:08:21

this DERP mechanism, D-E-R-P.

1:08:24

The only thing that's identifiable on there is, you know, like an

1:08:26

IP address, that kind of stuff, but frankly,

1:08:29

that's identifiable across many other

1:08:31

situations than just tail scale. The

1:08:33

actual packets, the actual traffic? Tail

1:08:36

scale aren't interested in that, because that would cost them money

1:08:38

to transit through their servers, all that kind of stuff. And

1:08:41

if you look at the free tier that tail scale offers too,

1:08:43

those particular users that are on the

1:08:45

free tier, you think to yourself, well, how

1:08:48

can that possibly make sense from a business perspective

1:08:51

for them to offer this thing for free? Well,

1:08:53

it's because no data transits tail scales network,

1:08:55

and they're still using something like tail scale funnel or something

1:08:57

like that, and it's relayed through a TCP proxy.

1:09:00

But that's from like hosting a very simple website as

1:09:03

a developer or something like that. So,

1:09:05

you know, in terms of trusting them, there isn't much

1:09:07

trust, they actually need to give tail scale.

1:09:09

The only thing I need to give them is that they're going to continue

1:09:12

to exchange my wire guard keys for me. And

1:09:14

that's it really, that's the only trust that I need to put

1:09:17

in tail scale as a company and

1:09:19

to continue to exist, of course. Yeah, and

1:09:21

I also probably could have done more to play

1:09:23

around with head scale. But part of

1:09:26

the overall solution that I find super attractive

1:09:28

is that the hosted component where they

1:09:30

do that sort of derp discovery, I guess, would

1:09:33

be the term, which I love, is

1:09:36

sort of the value added that I needed to actually

1:09:38

start using it. I kept having

1:09:40

one off wire guard setups that would

1:09:43

I would have for six months or nine months at

1:09:45

a time, maybe a year at most, then

1:09:47

I'd rebuild it and set up in a different way. And I have a different

1:09:49

way to, you know, maybe even try to manage

1:09:51

it. And

1:09:53

that just,

1:09:54

that worked. But it was, it's

1:09:56

like it's a, it's a, it's a, it's a tail scale is

1:09:58

a different level of a solution. Because it brings

1:10:01

all my devices into one mesh network

1:10:04

And I was never gonna get to that level with

1:10:07

anything that I had to kind of self set

1:10:09

up like that It's just it just what I was never gonna spend

1:10:11

the time And so that's kind of that

1:10:13

extra value they brought there is what made me put

1:10:15

it on everything You could argue that switching

1:10:17

solutions every six months is pretty secure

1:10:20

because by the time any script kiddies are cool Talk with

1:10:22

you you're on to the next thing. Yeah, that's true.

1:10:24

That's true Now we do have more boosted

1:10:26

in making to the show because we had ten total boosters

1:10:29

this week But we'll put a link to the boost barn. We

1:10:31

do read all of them, and then we save them for posterity

1:10:34

So thank you everybody who boosts in we stacked

1:10:36

three hundred and forty six thousand eight hundred and

1:10:38

four sats this week we really appreciate

1:10:40

that you can boost in by getting a podcast

1:10:43

app at podcast apps calm like pod verse

1:10:45

and Cast-a-matic and

1:10:47

fountain those are really popular in our audience and

1:10:50

if you don't want to switch apps We'll just get out beat get albie

1:10:52

calm you top it off and head over to the podcast

1:10:54

index and boost and we'll have links To that and

1:10:57

also thank you to our members once again. Don't forget. We got that

1:10:59

black Friday So two bucks off

1:11:01

a month for a year. We use promo code black

1:11:03

Friday You can

1:11:06

find me at Alex dot Katie said me

1:11:08

got a bunch of links to the various places But

1:11:11

I'm on the internet over there one

1:11:13

place you can find me as Linux unplugged Linux

1:11:15

unplugged calm

1:11:17

Oh bunches shenanigans happening over there,

1:11:19

too

1:11:20

Yeah, you can find me over there sometimes not

1:11:22

every week though sometimes. It's evil Chris You

1:11:24

just have to figure out if it's good Chris or evil Chris, but this

1:11:26

is good Chris for sure Although if it was

1:11:29

evil Chris I wouldn't tell you but they're both tweeting

1:11:31

at Chris LAS on weapon X and the show

1:11:33

is also at self hosted show if you Want to follow that

1:11:35

over there? Thanks for listening everybody. That

1:11:37

was self hosted not show