Bug 2203794

Summary: [ERFE] OVS needs a way to get new flow insertion/delete rate statistics.
Product: Red Hat Enterprise Linux Fast Datapath Reporter: Eelco Chaudron <echaudro>
Component: openvswitch3.1Assignee: Eelco Chaudron <echaudro>
Status: CLOSED DEFERRED QA Contact: ovs-qe
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: FDP 23.GCC: ctrautma, i.maximets, jhsiao, ralongi
Target Milestone: ---   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2023-11-24 12:28:19 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Eelco Chaudron 2023-05-15 09:31:45 UTC
It would be a good troubleshooting tool to get statistics related to flow insertion and delete rates.

This should exist for both the kernel and userspace datapath, preferably something unified. In addition, it might also be nice to not only get this globally but also for offloaded flows.

Maybe we can get some kind of histogram to get the various insertions/deletions per second. Like what currently exists for PMDs only:

  ovs-vsctl set Open_vSwitch . other_config:pmd-perf-metrics=true
  ovs-appctl dpif-netdev/pmd-perf-show

Comment 1 Ilya Maximets 2023-05-19 15:35:56 UTC
FWIW, there are coverage counters for dpif_flow_get/put/del/flush
and coverage/show provides the rate information.

Comment 2 Eelco Chaudron 2023-05-22 08:19:03 UTC
(In reply to Ilya Maximets from comment #1)
> FWIW, there are coverage counters for dpif_flow_get/put/del/flush
> and coverage/show provides the rate information.

ACK, but the put one only work for the kernel datapath, as OVS-DPDK pushes the flows directly in the PMD threads.

Comment 3 Eelco Chaudron 2023-11-24 12:28:19 UTC
CLosing, as https://issues.redhat.com/browse/FDP-158 was opened.