David Murphy
David Murphy
looks like hitting the race condition in process killing children (circa line 700) and not handling nicely
closing since PR https://github.com/saltstack/salt/pull/65937 is merged
problematic code (at the moment) appears to be here https://github.com/saltstack/salt/blob/66bb993cbd2103f00cbe40c13a89a9894c6ea1ed/salt/state.py#L2257-L2288 where not returning after the proc.start(), that is, code does not progress
The problem appears to be due a request for the cp.cache_file to the master getting lost since the parallel task has different pid etc, and gets lost in woodwork :(...
This appears to be an issue with the master only recognizing the originating pid, e.g. pid 26942, the parallel pid 26958, but all the master talks about is 26942, for...
the process manager kill_children has an empty self._process_map when called, but clearly it should not, except if you remove the parallel part of the state file, and re-run the command...
Will investigate this as part of the fixes for https://github.com/saltstack/salt/issues/65264, and PR https://github.com/saltstack/salt/pull/66218, since altering the files affected. Need to make sure ownership is preserved by these kinds of updates
Fixed with PR https://github.com/saltstack/salt/pull/66218
@gclinch Closing this since PR https://github.com/saltstack/salt/pull/66218 has been merged and should be available in 3006.9 which should be released very soon.
@dwoz Confirmed this is real with Ubuntu 22.04 on an AWS VM with Salt 3006.8, but checking libvirt on PyPI, ``` Requirements To build the libvirt python binary requires the...