[00:53:19] <Jhs>	 hi, i'm trying to migrate my telegram-wikilinksbot service from the old `jstart` stuff to the new `toolforge` stuff
[00:53:54] <Jhs>	 is this the right channel for this question, btw?
[00:54:52] <wm-bb>	 <jhsoby> i guess so, going off mw:Help:Toolforge
[00:55:41] <wm-bb>	 <jhsoby> okay, so i've set up a run.sh with the command i used to use before: `my_venv/bin/python3 wikilinksbot.py`
[00:56:38] <wm-bb>	 <jhsoby> this works when i just do `./run.sh` inside the ssh session, but it doesn't when i use it with `toolforge jobs run wikilinksbot --command ./run.sh --image bullseye --continuous`
[00:57:19] <wm-bb>	 <jhsoby> any idea what i'm doing wrong? should the command in `run.sh` have a more specific path, perhaps?
[01:09:17] <wm-bb>	 <lucaswerkmeister> what kind of error do you get?
[01:09:34] <wm-bb>	 <lucaswerkmeister> there should be a wikilinksbot.err file iirc
[01:10:15] <wm-bb>	 <lucaswerkmeister> oh wait – if it's Python you probably want a python-something image, the plain bullseye one might not be enough
[01:14:33] <wm-bb>	 <jhsoby> oh, there's a section here i hadn't noticed: https://wikitech.wikimedia.org/wiki/News/Toolforge_Stretch_deprecation#Rebuild_virtualenv_for_python_users
[01:14:37] <wm-bb>	 <jhsoby> will give that a try
[01:16:34] <wm-bb>	 <lucaswerkmeister> yeah, rebuilding is probably a good idea too
[01:16:58] <wm-bb>	 <lucaswerkmeister> (or just use a different venv / build it in run.sh maybe)
[01:22:21] <wm-bb>	 <jhsoby> Job name:     Job type:    Status:
[01:22:21] <wm-bb>	 <jhsoby> ------------  -----------  ------------------------------
[01:22:23] <wm-bb>	 <jhsoby> wikilinksbot  continuous   Specified command fails to run
[01:22:37] <wm-bb>	 <jhsoby> (but again, just run.sh on its own works as expected)
[01:55:11] <wm-bb>	 <jhsoby> alright, i found the actual error in the .err file:
[01:55:12] <wm-bb>	 <jhsoby> ModuleNotFoundError: No module named 'telegram'
[01:55:14] <wm-bb>	 <jhsoby> 
[01:55:15] <wm-bb>	 <jhsoby> Even though it's supposed to be running inside the virtual environment specifically created to install that module 🤔
[11:01:39] <wm-bb>	 <lucaswerkmeister> I’m confused by how the `run.sh` even reaches any Python code if the command is `.././venv/bin/python3`
[11:02:07] <wm-bb>	 <lucaswerkmeister> I would expect the working directory to be the tool home directory, in which case that path doesn’t resolve
[11:06:57] <taavi>	 @jhsoby: python virtual envs are linked to the environment in which they were created in. so if it works on a bastion, that usually means it was created on a bastion and won't work in a kubernetes-based job. see https://wikitech.wikimedia.org/wiki/Help:Toolforge/Python#Virtual_environments
[11:07:00] <wm-bb>	 <lucaswerkmeister> which image did you end up using? I can’t see it anywhere
[11:07:23] <taavi>	 the 'ModuleNotFoundError' error is also a sign that the venv is being used in an environment it was not created in
[11:08:13] <wm-bb>	 <jhsoby> I tried with both python images too, same result (re @lucaswerkmeister: which image did you end up using? I can’t see it anywhere)
[11:09:13] <wm-bb>	 <lucaswerkmeister> I’m assuming you mean python3.9 and python3.11? (those are the ones I can see in `toolforge jobs images`)
[11:09:17] <wm-bb>	 <jhsoby> Yeah, i created a new venv to test with, but I probably did it incorrectly (re @wmtelegram_bot: <taavi> @jhsoby: python virtual envs are linked to the environment in which they were created in. so if it works on a bastion, t...)
[11:09:19] <wm-bb>	 <lucaswerkmeister> neither of those could work, the venv is a python3.7 one
[11:09:25] <wm-bb>	 <jhsoby> Yes (re @lucaswerkmeister: I’m assuming you mean python3.9 and python3.11? (those are the ones I can see in toolforge jobs images))
[11:10:09] <wm-bb>	 <jhsoby> Makes sense. I also noticed that, but at that point it was too late for me brain to fully process it 😜 (re @lucaswerkmeister: neither of those could work, the venv is a python3.7 one)
[11:11:48] <wm-bb>	 <lucaswerkmeister> I think creating the venv in run.sh is probably the safest way to ensure it’s compatible, I do something similar in the quickcategories background runner https://github.com/lucaswerkmeister/tool-quickcategories/blob/d817ca5f60ba3d3ae9303acf0eced31caabfe231/deployment.yaml#L29
[11:12:10] <wm-bb>	 <lucaswerkmeister> but using a `webservice shell` probably also works as long as you use the kubernetes backend and the right webservice type (with the right python version)
[13:16:51] <wm-bot>	 !log tools.lexeme-forms <lucaswerkmeister> deployed 442f510a5b (German neuter noun Wikifunctions)
[13:16:54] <stashbot>	 Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools.lexeme-forms/SAL
[15:04:14] <wm-bot>	 !log tools.lexeme-forms <lucaswerkmeister> deployed 11e7d12745 (one more set of German neuter noun Wikifunctions)
[15:04:18] <stashbot>	 Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools.lexeme-forms/SAL
[16:36:17] <wm-bot>	 !log tools.lexeme-forms <lucaswerkmeister> deployed 0416376e58 (German masculine noun Wikifunctions)
[16:36:21] <stashbot>	 Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools.lexeme-forms/SAL
[20:25:55] <wm-bot>	 !log tools.pbbot <peterbowman> Restart webservice to solve 'no class def found' error
[20:25:58] <stashbot>	 Logged the message at https://wikitech.wikimedia.org/wiki/Nova_Resource:Tools.pbbot/SAL
[23:54:29] <wm-bb>	 <jhsoby> i'm trying a similar approach now, and again it works when i just run ./run.sh, but fails (with "Specified command fails to run") when i do it with `toolforge jobs run` 😩 (re @lucaswerkmeister: I think creating the venv in run.sh is probably the safest way to ensure it’s compatible, I do something similar in the quickcat...)
[23:54:57] <wm-bb>	 <jhsoby> so maybe my paths are wrong?
[23:57:29] <wm-bb>	 <lucaswerkmeister> it looks like you’re still using the `bullseye` image, it should probably be one of the python ones