8mo ago

2024-07-18 lemm.ee downtime

Hey folks!

Unfortunately, roughly 2 hours ago, lemm.ee went offline. The cause was our load balancer: it suddenly decided that all of our servers had become unhealthy, despite all health checks responding successfully when I requested them directly. In such cases, the load balancer stops serving all requests, effectively meaning that lemm.ee is unreachable for all users. I am still not sure what exactly caused the issue, but I will try to investigate more over the weekend.

For now, we have partially recovered, and I am continuing to work on remaining issues. Hopefully we will be back to 100% very soon. Sorry for the inconvenience!

29 comments

We appreciate what you do hero

I was wondering what was going on, status.lemm.ee said the server was ok but the federation was broken. Thank you for fixing it
- Sorry for the delay in updating the status page - I actually had gone out for lunch just a few minutes before the downtime started, so I didn't even realize anything was up until I was back at my computer about 45 minutes later 💀
  
  no need to apologise. still a better response time, than some of the professionals I work with ;-)

I survived the July 18th lemm.ee downtime, and all I got was this lousy comment.

All is forgiven, thank you for running this lovely instance ^_

Thanks for your great work and transperancy!

I'd like to speak to a manager /s

Nginx? I had an nginx LB shit itself yesterday. Luckily it auto-recovered and I had HA but just weird it happened.
- Actually, we're using Hetzner's cloud load balancer for lemm.ee. But if this issue repeats in the near future, then I will definitely consider setting up something else.
  
  haproxy is where it's at!

Thanks for the quick fix! What did you have to do to get the load balancer working again?
- For now, I just redeployed all of our servers completely, but as I don't know the actual root cause of the issue yet, I'm still investigating to figure out if anything more is needed.

Typically when this happens, the issue is on the LB itself. Maybe its own network had issues?

Sometimes, downtimes are awesome. Get off your machine and spend time with your family, folks!

Would it be in bad taste to blame Russia?
- Yeah, but it could have been China, India, Iran, or maybe even North Korea. There are a lot of places that think disrupting the rest of the world will get them somewhere.

I thought the entire lemmy network was down because status.lemm.ee was saying our instance was fine and federation wasn't working with every other instance. lol

love you guys!

Seriously, your professionalism in handling the situation and in reporting it is fantastic.
It's totally above and beyond anything we should expect for a service powered by donations!
Thank you!

Is there another instance where you could report issues?
If we logged into another account, we’d be able to see those before it comes back up.
- There are two useful sections on https://status.lemm.ee for this - firstly, there is an automated check for federation with all other instances on the bottom of the page, and everything there being red is a definite sign that something is wrong with lemm.ee itself. Secondly, near the top of that page, I will always write a status message manually when I discover & start work on any issues. This second part can have a bit of a delay, as it requires manual input from myself, but I have updated it every time we had any issues so far.
  
  That’s good info. Thanks.
- There's a Discord server in the sidebar that updates are posted in: https://discord.gg/XM9nZwUn9K
  
  I know. I just don’t want to join a discord.

Thank you for keeping us abreast of what's happening. I appreciate you, and how you manage this instance.

Totally healthy servers have a right to rest every once in a while too. Thanks for keeping us notified!

29 comments