Unified agent host charm #437
Conversation
* Move run_with_logged_errors out of charm.py * Update testflinger source and add action for it * Add integration tests and docs for the update-testflinger action * cleanup
* Write out the supervisord configs for all agents * Update the supervisor services when configs/code is updated * Update README for the agent charm with instructions for test/debug * redirect stderr for all supervisord agents * Fix charm to expect base64 encoded ssh keys as already documented * important fix - reservations won't work without this Any script (like ssh-copy-id) which uses $HOME, or ~ or anything like that won't work without this. Although supervisord execs the process as the specified user, there's no shell. So we need to set certain vars like this in the supervisord config. * Also install the device-connector so that things like reserve work * Rename ssh config options for better consistency * Update testflinger-agent-host-charm README
* Add agent-host terraform module * Separate variables from main.tf
* Fix some formatting issues in the README.md * Add variables to build constraints to the agent-host terraform module
There was a problem hiding this comment.
Super great work! I appreciate the benefit brought by the unified agent host charm :)
I only have a small suggestion.
Plus, it would be nice to resolve the open files limit of supervisorctl that we've discussed. Below is the limit I put in supervisord.conf. As experimented, 336 agents would exceed the default limit of 1024. So it can be either lower than 65535 or we can just put it as a safe number.
[supervisord]
minfds=65535
agent/charms/testflinger-agent-host-charm/src/tf-cmd-scripts/tf-cleanup
Outdated
Show resolved
Hide resolved
|
@nancyc12 for the minfds thing, I'm looking at doing that as a separate commit if that's ok. There's some other fixes/cleanups I'd like to do first, but I figured we could land this one for now since it's well tested at this point. Also it won't revert any change to the supervisord.conf that you made even if you deployed it |
|
@nancyc12 I also pushed a few more changes to fix some test issues |
Description
We've actually been running this for a while now, but have held off on landing it because the old version of the charms were running in the devices lab still. It looks like it should be safe to merge now though!
Resolved issues
CERTTF-374
Documentation
Updated READMEs and deployment instructions
Web service API changes
N/A
Tests
This is currently used in our labs and has been running smoothly for a while now.