Closed dmick closed 2 years ago
@djgalloway fyi. I think I want to discuss the last commit with @andrewschoen before changing this out of draft
@djgalloway fyi. I think I want to discuss the last commit with @andrewschoen before changing this out of draft
I took a look at https://github.com/ceph/chacra/pull/301/commits/ea1a489a28537f88d09e9b5ede552c2ddaca0835
They only reason for using librabbitmq
that I remember is because the docs recommended it. Reading through the ticket you mentioned in the commit and this confirms that we should not be using librabbitmq
anymore: https://github.com/celery/celery/pull/6933/commits/38b0527b170408f1d86bd48890fa7e47a35f9baf
Added another commit that had been on another pending branch to fix binary checksumming; this was preventing binary uploads. It really should have been part of the python3 port, but the code wasn't exercised until I got a test instance up and running, and then I probably should have put it on a different branch, but it was on the wip-add-raw-repos branch because that's where I tested it.
Anyway, it's on this branch now and will resolve the current issue with 1.chacra.ceph.com.
1) really fix the checks.py call for celery_has_workers() (not Inspect(), but app.control.inspect()) 2) fix up disk_has_space checks: wrong config item name, popen() result not waited for, Py3 stdout/err are binary 3) force use of py-amqp rather than librabbitmq; the latter is apparently deprecated and is causing problems. 4) improve disk_has_space: check both repos_root and binary_root 5) (added late): fix binary checksum calculation for Py3.