[counters]: separate query of port/queue counters #483
Conversation
Signed-off-by: Sihui Han <[email protected]>
| @@ -34,8 +34,10 @@ extern CrmOrch *gCrmOrch; | |||
|
|
|||
| #define VLAN_PREFIX "Vlan" | |||
| #define DEFAULT_VLAN_ID 1 | |||
| #define FLEX_STAT_COUNTER_POLL_MSECS "1000" | |||
| #define STAT_COUNTER_FLEX_COUNTER_GROUP "STAT_COUNTER" | |||
| #define PORT_FLEX_STAT_COUNTER_POLL_MSECS "1000" | |||
There was a problem hiding this comment.
Are both polling intervals controlled by cli or through fields in config_db.json?
If yes, can you please tell me how?
If not, then those intervals need to have more control.
There was a problem hiding this comment.
Also queue counters polling must not be enabled by default.
There was a problem hiding this comment.
@Nikos-Li , please create feature request in the issue, we'll triage later.
There was a problem hiding this comment.
@lguohan This isn't a feature request, nor a proper solution to the CPU increase we have observed. This needs to be discussed in the 3-way meeting. Please include it in the agenda.
There was a problem hiding this comment.
@lguohan @sihuihan88 Please also provide CPU util data of before and after and a proper root cause analysis as well as the testing done, verification etc. None of this exists in the PR and an almost 200% increase in CPU util, makes this a critical fix.
There was a problem hiding this comment.
@lguohan @sihuihan88 Do not merge this without proper RCA and explicit approval and discussion in the 3-way meeting.
Signed-off-by: Sihui Han <[email protected]>
Signed-off-by: Sihui Han [email protected]
What I did
Separate port and queue query. By default query port counters every 1s, and queue counters every 10s
Why I did it
Querying all queue counters every 1s is cpu consuming. Reducing queue counter query rate helps to reduce cpu overhead.
How I verified it
Tested on DUT.
Details if related