Bug #43402: Python 3.6 + ceph_argparse.py RadosThreads may not fully exit before is_alive() is called - Ceph - Ceph

Actions

Copy link

Bug #43402

closed

Python 3.6 + ceph_argparse.py RadosThreads may not fully exit before is_alive() is called

Added by Mark Nelson over 4 years ago. Updated almost 4 years ago.

Status:

Resolved

Priority:

Normal

Assignee:

Mark Nelson

Category:

ceph cli

Target version:

v15.0.0

% Done:

Source:

Development

Tags:

Backport:

Regression:

Yes

Severity:

3 - minor

Reviewed:

Affected Versions:

ceph-qa-suite:

Pull request ID:

32379

Crash signature (v1):

Crash signature (v2):

Description

Running the ceph command from a recent pull of master where python 3.6 is used always results in an exception being thrown on one of our incerta test boxes:

[perf@incerta01 log]$ ceph -w
Traceback (most recent call last):
  File "/usr/local/bin/ceph", line 1267, in <module>
    retval = main()
  File "/usr/local/bin/ceph", line 980, in main
    conffile=conffile)
  File "/usr/local/lib/python3.6/site-packages/ceph_argparse.py", line 1336, in run_in_thread
    raise Exception("timed out")

This appears to be the result of calling is_alive() immediately after join where a thread hasn't fully finished exiting yet. On my system inserting a short sleep is enough to resolve the issue, though this may not resolve the situation on all platforms:

time.sleep(.01)

A better option might be a short loop with multiple sleeps and retries of is_alive() before finally giving up. Someone who knows python better than I do may have a superior solution. As it is, this appears to block all usage of the ceph command.