yahoo-eng-team team mailing list archive
-
yahoo-eng-team team
-
Mailing list archive
-
Message #95839
[Bug 2110085] [NEW] neutron-api: does not work with multiple uwsgi processes and threads
Public bug reported:
Hi,
It seems this is limited to when using OVN, though I'm still forwarding this Debian bug:
https://bugs.debian.org/1093303
As per the reporter:
the service nova-api starts it writes a lock in ovn_hash_ring table in
the neutron database. With multiple processes or threads from uwsgi it
does not update this record anymore. After a while when running
"openstack network agent list" it will yield an error:
ERROR neutron.api.v2.resource [None req-b2f34437-1e6f-417f-86b9-e4e21f2b3d37 119654c86dd74332972dc6528e5af420 9e18eb1ee7cc423a88cd44e901ecc119 - - default default] index failed: No details.:
ovsdbapp.exceptions.TimeoutException: Commands [CheckLivenessCommand(_result=None)] exceeded timeout 180 seconds, cause: Result queue is empty
Setting a single thread and a single process in the uwsgi .ini file
fixes the proble, however, as you may know, this isn't really usable in
production, where dozens of workers are needed.
Please get this fixed.
** Affects: neutron
Importance: Undecided
Status: New
--
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/2110085
Title:
neutron-api: does not work with multiple uwsgi processes and threads
Status in neutron:
New
Bug description:
Hi,
It seems this is limited to when using OVN, though I'm still forwarding this Debian bug:
https://bugs.debian.org/1093303
As per the reporter:
the service nova-api starts it writes a lock in ovn_hash_ring table in
the neutron database. With multiple processes or threads from uwsgi it
does not update this record anymore. After a while when running
"openstack network agent list" it will yield an error:
ERROR neutron.api.v2.resource [None req-b2f34437-1e6f-417f-86b9-e4e21f2b3d37 119654c86dd74332972dc6528e5af420 9e18eb1ee7cc423a88cd44e901ecc119 - - default default] index failed: No details.:
ovsdbapp.exceptions.TimeoutException: Commands [CheckLivenessCommand(_result=None)] exceeded timeout 180 seconds, cause: Result queue is empty
Setting a single thread and a single process in the uwsgi .ini file
fixes the proble, however, as you may know, this isn't really usable
in production, where dozens of workers are needed.
Please get this fixed.
To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/2110085/+subscriptions