supervisor: Do not nuke nodes after jobs finish

This was causing a bad race condition, where we could unlock a node, then unlock
it again via the nuke process after a different job had locked it.

Fixes: https://tracker.ceph.com/issues/64193
Signed-off-by: Zack Cerza <zack@redhat.com>

diff --git a/teuthology/dispatcher/supervisor.py b/teuthology/dispatcher/supervisor.py
index 6af6f554a0ea80eb0a8f731a6ac705afc2a9048a..bb3885d74389788e4c4b213c8a43a3ef1642aca1 100644 (file)
--- a/teuthology/dispatcher/supervisor.py
+++ b/teuthology/dispatcher/supervisor.py
@@ -282,10 +282,6 @@ def unlock_targets(job_config):
                 machine, job_info['owner'],
                 job_info['archive_path']
             )
-    if job_status != 'pass' and job_config.get('nuke-on-error', False):
-        log.info('Nuking machines...')
-        fake_ctx = create_fake_context(job_config)
-        nuke.nuke(fake_ctx, True)
 
 
 def run_with_watchdog(process, job_config):