Page MenuHomePhabricator

[bug] PAWS has problems loading
Closed, ResolvedPublic

Description

What happened?

PAWS ran smoothly until about an hour ago. Started with having a hard time getting the kernel to do anything, after which reloading resulted in 504 errors and blank screens.

I'm getting this error in the js console (and lots of similar errors):
Unsatisfied version 3.4.8 from @jupyterlab/application-top of shared singleton module @jupyterlab/apputils (required ^3.6.3)

Using Chrome on Mac. When I delete my stored data for wmcloud.org and am logged out, I do see the login screen and can log in. Immediately after logging in through mediawiki.org I get the same problems. Eventually it will load after several minutes, but then state
Server unavailable or unreachable
Your server at /user/Effeietsanders/ is not running. Would you like to restart it?

Event Timeline

Found a workaround: stopping the server via the hub control panel, and then starting a server from the same resulted in a workable situation again. Not sure if this should be closed, or if there's an underlying issue that needs addressing.

This might be related to T338981. Jupyterlab is back versions at the moment as described in that ticket. Though keeping jupyterlab back is causing some limitations in other areas, this ticket being one. Let's keep this open for now.

@rook just fyi, I was having problems again today (between 9 and 10pm PT, if that helps) as a gradual slowdown and restarting kernel didn't help, but this was resolved after restarting the server.

Hi @Effeietsanders how long does it take for you to see a slowdown after you start your server? Or perhaps after you start working with the code you're using? Could you link me to the code that stops?

I have been running different versions of two notebooks (iterating to squeeze out bugs and data issues) this and this. I suspect the latter has been the most challenging - but it's hard to pin down as it only happened twice so far.

@Effeietsanders could you see if you can recreate the issue on hub-paws-dev.codfw1dev.wmcloud.org ?

@rook I've been unable to reproduce even on the main server, but the hub-paws-dev actually doesn't even start up the server for me (it gets stuck at 2023-08-01T23:22:42Z [Normal] Pulling image "quay.io/wikimedia-paws-prod/singleuser:pr-317")

@rook I've been unable to reproduce even on the main server, but the hub-paws-dev actually doesn't even start up the server for me (it gets stuck at 2023-08-01T23:22:42Z [Normal] Pulling image "quay.io/wikimedia-paws-prod/singleuser:pr-317")

Oh I think that was because I forgot to do an initial login, so yours was the initial login. The singleuser container is big, so it usually takes longer than the timeout to get pulled down. Though the first login initiates the pull, so by the second login it is available. That said, I guess it doesn't make much difference what is happening in dev if the bug cannot be seen in production.

T343116 fixed some problems related to jupyterlab. I'm curious if this behavior can still be observed?

Sounds like this might have been resolved. Please re-open if not found to be the case.