[neutron][largescale-sig] agent_down_time and report_interval
Hello neutron and large-scalers, Is there any recommendation on tuning the neutron report_interval (agent) and agent_down_time (server) to "optimize" the communication between agents and servers without putting to much heavy duty on both rabbit and database? We are currently facing some scaling issue regarding this, and we found out that, at least, CERN did some tweak about this ([1] and [2]) Is there anyone else with specific configuration on that part? I have the feeling that this could be increased (so report will happen less often). One obvious side effect of this is the fact that the server will take more time to see a down agent, is there any other side effect that could happen? Is it something we can eventually add in our large-scale documentation ([3])? Cheers, [1] https://youtu.be/5WL47L1P5kE?t=1173 [2] https://object-storage-ca-ymq-1.vexxhost.net/swift/v1/6e4619c416ff4bd19e1c08... [3] https://docs.openstack.org/large-scale/journey/configure/index.html
Hi everyone, i can share that we use a agent_down_time of 3600 seconds and the default report_interval for clusters with around 600 compute nodes. -- Felix Huettner
-----Original Message----- From: Arnaud Morin <arnaud.morin@gmail.com> Sent: Wednesday, March 8, 2023 2:30 PM To: discuss openstack <openstack-discuss@lists.openstack.org> Subject: [neutron][largescale-sig] agent_down_time and report_interval
Hello neutron and large-scalers,
Is there any recommendation on tuning the neutron report_interval (agent) and agent_down_time (server) to "optimize" the communication between agents and servers without putting to much heavy duty on both rabbit and database?
We are currently facing some scaling issue regarding this, and we found out that, at least, CERN did some tweak about this ([1] and [2])
Is there anyone else with specific configuration on that part?
I have the feeling that this could be increased (so report will happen less often). One obvious side effect of this is the fact that the server will take more time to see a down agent, is there any other side effect that could happen?
Is it something we can eventually add in our large-scale documentation ([3])?
Cheers,
Diese E Mail enthält möglicherweise vertrauliche Inhalte und ist nur für die Verwertung durch den vorgesehenen Empfänger bestimmt. Sollten Sie nicht der vorgesehene Empfänger sein, setzen Sie den Absender bitte unverzüglich in Kenntnis und löschen diese E Mail. Hinweise zum Datenschutz finden Sie hier<https://www.datenschutz.schwarz>.
Thanks! On 08.03.23 - 13:58, Felix Hüttner wrote:
Hi everyone,
i can share that we use a agent_down_time of 3600 seconds and the default report_interval for clusters with around 600 compute nodes.
-- Felix Huettner
-----Original Message----- From: Arnaud Morin <arnaud.morin@gmail.com> Sent: Wednesday, March 8, 2023 2:30 PM To: discuss openstack <openstack-discuss@lists.openstack.org> Subject: [neutron][largescale-sig] agent_down_time and report_interval
Hello neutron and large-scalers,
Is there any recommendation on tuning the neutron report_interval (agent) and agent_down_time (server) to "optimize" the communication between agents and servers without putting to much heavy duty on both rabbit and database?
We are currently facing some scaling issue regarding this, and we found out that, at least, CERN did some tweak about this ([1] and [2])
Is there anyone else with specific configuration on that part?
I have the feeling that this could be increased (so report will happen less often). One obvious side effect of this is the fact that the server will take more time to see a down agent, is there any other side effect that could happen?
Is it something we can eventually add in our large-scale documentation ([3])?
Cheers,
Diese E Mail enthält möglicherweise vertrauliche Inhalte und ist nur für die Verwertung durch den vorgesehenen Empfänger bestimmt. Sollten Sie nicht der vorgesehene Empfänger sein, setzen Sie den Absender bitte unverzüglich in Kenntnis und löschen diese E Mail. Hinweise zum Datenschutz finden Sie hier<https://www.datenschutz.schwarz>.
participants (2)
-
Arnaud Morin
-
Felix Hüttner