Details
-
Feature Request
-
Resolution: Unresolved
-
Undefined
-
None
-
None
-
False
-
None
-
False
-
Not Selected
-
0
-
0%
Description
1. Proposed title of this feature request
Collect openvswitch metrics.
2. What is the nature and description of the request?
Scrape OVS counters into Prometheus an example of some counters is below.
[root@worker-0 quicklab]# ovs-dpctl show -s system@ovs-system: lookups: hit:74256356 missed:3464462 lost:0 flows: 398 masks: hit:644338083 total:54 hit/pkt:8.29 port 0: ovs-system (internal) RX packets:0 errors:0 dropped:0 overruns:0 frame:0 TX packets:0 errors:0 dropped:0 aborted:0 carrier:0 collisions:0 RX bytes:0 TX bytes:0 port 1: ens3 RX packets:21575002 errors:0 dropped:0 overruns:0 frame:0 TX packets:11753410 errors:0 dropped:0 aborted:0 carrier:0 collisions:0 RX bytes:19301239828 (18.0 GiB) TX bytes:10957854861 (10.2 GiB) port 2: br-ex (internal) RX packets:13694407 errors:0 dropped:0 overruns:0 frame:0 TX packets:9989227 errors:0 dropped:0 aborted:0 carrier:0 collisions:0 RX bytes:17052774759 (15.9 GiB) TX bytes:10648742168 (9.9 GiB)
3. Why does the customer need this? (List the business requirements here)
Customers currently have no insights into the number of upcalls/flows getting created inside their clusters, This results in a lack of awareness.
It would greatly benefit support and customers if we could monitor the number of flows and hit/miss ratios. and would allow for insights to also take advantage in the future.
Mainly we need some way to check for excessive `upcalls` cause by excessive flow eviction.
4. List any affected packages or components.
OVS/Monitoring
