General Lemmy.Cafe

Sorry for causing all the inconvenience

Everything went through smoothly, but please do report if you notice anything.

Lemmy Cafe will be having its database upgraded.

Reasons

• PostgreSQL 17 has been released and the changelog is promising a lot of IO improvements. Lemmy sure could use it given the constant stream of small events flowing in.

Plan

• Point nginx to the maintenance page
• Shut down PostgreSQL 16
• Run the upgrade tool
• Start up PostgreSQL 17
• Point nginx to lemmy

Expected downtime

About an hour, if things go well. More if not so.

Will try to keep the maintenance page updated.

Here's the timezone converter.

Due to the recent @Soup's post I have decided to do some housekeeping. I've been getting frustrated at lemmy's performance at times as well and this simply was a wake up call, if you will.

I am sorry about no advance downtime annoucement - Sunday is the only day I can really put any meaningful amount of time into lemmy.

___ Things done today:

• Upgraded database VM 1 core 2GB -> 2 cores 4GB. Double the compute, double the memory.
• Adjusted database config to account for increased resources
• Adjusted hugepages config to account for increased database's requests
• Updated both lemmy and database VMs
• Rebooted the lot

___

Thank you for your patience. I will also use this moment of focus to write up a financial report in a separate post.

Also, thank you, @Soup!

___ EDIT: I've also marked all instances known to lemmy.cafe as active. What this means is that lemmy.cafe will now keep retrying to federate to everything very aggressively. This is a compute-intensive process and will also impact performance for a few hours, until exponential backoff kicks in. I've done it to revive any falsely-marked-as-dead instances; there's no fix for it on lemmy itself.

Everthing about this instance crawls.

We're now running on dual thread 4 gigs or ram. Hopefully that alleviates some recent instability.

Database is on a separate VM, seems to be holding up so far.

This was a bugfix release.

The !general@lemmy.cafe has been reverted to being public as the private switch appears to break notifications.

Due to increased compute demand of lemmy software the VM lemmy.cafe is running on had to become a little beefier. DB VM is still chugging along without changes.

Happy socials!

___

As always - any financial help is greatly appreciated! To donate click the banner here or in the sidebar:

[!](https://liberapay.com/Illecors/donate)

❤, Illecors

Timeout uploading images, either for posts or comments. Tried Lemmy-UI, Voyager and Photon.

This is a big release - the most interesting, probably, being local [non-federated] communities. There are a lot of other features, as well, so don't hesitate reading the release announcement!

As a first attempt, I have made !general@lemmy.cafe a local community. This post should not federate anywhere else.

The upgrade itself was a bit rocky, and, admittedly, caused by the setup that has strayed off the official guidelines. It has, however, helped in troubleshooting countless times previously, so it's not going away. More care will simply have to be taken next time.

For those curious - the bit that caused all the trouble was having separate users to connect to the database for each of backend, federation and tasks containers. Something changed where a new schema was created/started getting used and tasks container did not have access to said schema. As such, it could not finish the migration scripts.

As always - any financial help is greatly appreciated! To donate click the banner here or in the sidebar:

[!](https://liberapay.com/Illecors/donate)

❤, Illecors

example

unless there is an intentional reason for this, would love to see this working as i value the conversations going on over there :)

It's been a while since I've last done one of those. Life happened, but things are settling down slowly. I'll combine all five months since September into a single post, but will keep them separated visually.

Income is applied on date of arrival. Stripe (LiberaPay uses it as processor) pays out in advance and as such income shows up as 0, even though LiberaPay banner shows a non-zero value.

Any financial help is greatly appreciated! To donate click the banner here or in the sidebar:

[!](https://liberapay.com/Illecors/donate)

February 2024

Contributions

• LiberaPay: £0

Expenses

• Servers: £38.94

Month balance

• -£38.94

Previous balance

• -£216.05

Balance to date

• -£254.99

___

January 2024

Contributions

• LiberaPay: £0

Expenses

• Servers: £38.66

Month balance

• -£38.66

Previous balance

• -£177.39

Balance to date

• -£216.05

___

December 2023

Contributions

• LiberaPay: £0

Expenses

• Servers: £38.59
• Backblaze B2: £0.64

Month balance

• -£39.23

Previous balance

• -£138.16

Balance to date

• -£177.39

___

November 2023

Contributions

• LiberaPay: £0

Expenses

• Servers: £38.74

Month balance

• -£38.74

Previous balance

• -£99.42

Balance to date

• -£138.16

___

October 2023

Contributions

• LiberaPay: £0

Expenses

• Servers: £37.32

Month balance

• -£37.32

Previous balance

• -£62.1

Balance to date

• -£99.42

___

Finance History

| | September 2023 | August 2023 | July 2023 | June 2023 | |---------------|---------|---------|---------|---------| | Contributions | £11.41 | £15.26 | £0 | £0 | | Expenses | £28.85 | £24.75 | £19.18 | £15.99 | | Difference | -£17.44 | -£9.49 | -£19.18 | -£15.99 | | Balance | -£62.1 | -£44.66 | -£35.17 | -£15.99 |

___

Previous reports

___ September 2023 ___ August 2023 ___ July 2023 ___ June 2023

due to the offending user being banned from the federating instance before the post was removed? example

edit: i recognize this is not a local issue and may be more of a lemmy problem. wondering if anyone has any experience or thoughts cuz its weird to have awful stuff hosted on one instance even though it was banworthy elsewhere.

edit 3: removed edit 2.

Recently I was trying to reply to a comment and it wouldn't post the reply. Today I also wanted to upload a meme but the image wouldn't upload, so I resorted to using catbox.moe. Even with that, it still insisted on generating its own broken image and the thumbnail didn't generate either

UPDATE: Apparently only GIFs and replies are broken. Uploading a JPG or PNG works fine

https://lemmy.cafe/post/3646439

It's the "silently" part that's the issue. I acknowledge that lemmy.cafe is entitled to defederate from whatever servers the administration pleases, but lemmy.ml still houses some of the largest communities in the Lemmyverse on some topics, and a heads-up that it was being blocked would have been appreciated.

The upgrade has gone through smoothly and everything seems to be running smoothly.

The performance looks to be better on the backend, time will tell if the memory leak issue is actually solved. So far, though - so good!

Lemmy Cafe will be having its database upgraded.

Reasons

• Pict-rs is expecting PostgreSQL 16. It's running fine now, but it might not be at some future point.
• PostgreSQL 15 has a bug that requires jit to be turned off - otherwise DB keeps consuming all the memory available on the system and then some. In the end it gets culled by the kernel. This has performance as well as reliability issues. While turning jit off has remedied the constant failure, it has also made the database a bit slower. I prefer squeezing out as much performance as possible :)

Plan

• Point nginx to the maintenance page
• Shut down PostgreSQL 15
• Run the upgrade tool
• Start up PostgreSQL 16
• Point nginx to lemmy

Expected downtime

About an hour, if things go well. More if not so.

Will try to keep the maintenance page updated.

Here's the timezone converter

The process went through smoothly. I have also used the opportunity to split up a singular lemmy container into individual tasks - this has enabled a seemless upgrade process with no downtime, bar a few process quirks I need to work out.

There have been some federation fixes merged into this release, so the situation should definitely be improving overall!

I will make a more detailed write up of the whole setup later on, other admins might find it useful. Or not.

May your families grow strong and your presents plentiful!

The upgrade has reset the JWT, so you'll have to login again on your device as well as set up 2FA again.

Enjoy the new features!

Pun intended 😁

Just wanted to reiterate, my stance hasn't changed since last declaration.

If facebook starts funny business with extra domains - those will be added asap.

It's been a long one, but finally the database is clean. In the end the cleanup took <20 minutes, but getting to it took me most of the weekend.

Sorry about flaky performance, it should be back to normal.

Migrated email into a dedicated self-managed solution, but it was at the end of the month, haven't included it here.

Income

• LiberaPay: £11.41 ___ Expenses
• Servers: £28.85 ___ Month balance
• -£17.44 ___ Previous balance
• -£44.66 ___ Balance to date
• -£62.1

Previous reports

___ August 2023 ___ July 2023 ___ June 2023

Minor update in prep for the biggin. Pictrs was also pushed up to 0.4.4

The main feature being a fix the really nasty libwebp vulnerability.

I've just received an email that has links to what looks like some polish url redirect site. I suspect this is limited to admins as, by default, lemmy does not come with an email service.

Needless to say - don't click random links!

Stay safe, peeps!

I'm chuffed to bits to have finally had enough time to set it up externally. Google is no longer outright rejecting us!

And by externally I don't mean succumbing to the behemoths - it's all self managed!

I've seen a post recently about someone struggling hosting Lemmy on DigitalOcean and sending out verification emails. If you need help with config or would like to simply use lemmy.cafe's setup - feel free to reach out!

First donations!

Also migrated pictrs from file to object storage, but have not been invoiced, yet. I suspect it's the low amount (<$1) and will probably keep carrying over until some threshold is reached.

Income

• LiberaPay: £15.26 ___ Expenses
• Servers: £24.75 ___ Month balance
• -£9.49 ___ Previous balance
• -£35.17 ___ Balance to date
• -£44.66

Signed up on LiberaPay. Stripe is slow to activate the account.

Lemmy storage was growing quite quickly on the small VPS nodes, had to upscale a bit.

Income

• None ___ Expenses
• Servers: £19.18 ___ Previous balance
• -£15.99 ___ Balance to date
• -£35.17

Decided to finally do a write up financials.

Income

• None ___ Expenses
• Domain: £8.46
• Servers: £7.53 ___ Balance to date
• -£15.99

Sorry about it, it seems like after having nuked the CSAM incident community lemmy-ui has freaked out again about the site logo. Fixed it for now, will keep a close eye on the web UI.

Apps were not affected.

Following the recent CSAM spam the upload size has been further reduced to 50kB.

This might impact community/profile banners, in which case I propose getting in touch and we'll sort it out on a case by case basis.

Apologies for the downtime.

DigitalOcean has been experiencing a network outage and as such Lemmy Cafe is also affected.

Incident page suggests it should be resolved now.

Current infrastructure costs are ~30 GBP a month. Not crazy expensive, but something that has potential to grow in the future.

Expense breakdown:

• Lemmy UI+backend: 14GBP/month
• Lemmy DB: 14 GBP/month
• Domain: ~2 GBP/month
• Traffic is not reaching the limits at the moment, no extra cost

LiberaPay link

It is also present in the sidebar.

Everything appears to have gone through quite well, although the ambiguous UI error during the database migration was confusing.

Storage was getting full, it was time to upgrade.

https://feddit.de/c/ich_iel and https://lemmy.cafe/c/ich_iel@feddit.de show completely different posts.

Shouldn't both sites have the same posts?