Do not retry immediately after server fault
Currently the call_and_ignore_notfound_exc method retries API request
immediately after it receives server failure. However this is not much
robust because retries might be done in a quite short span.
For example in the related bug we've seen that glance-api with cinder
backend needs some time to complete the post tasks to disconnect
the volume when it's run by httpd + mod_wsgi. Currently the 3 delete
calls are executed within 1 second and that is not enough to complete
that task.
This adds 1 second delay between reties, assuming longer delay would
help the server stabilize.
Related-Bug: #2030855
Change-Id: Ic9eed8352c64f183ca4077e141c6b94f81a9b91a
1 file changed