[PFCWD]: Periodically poll WD counters #473
Conversation
For every action handler call commitCounters on a periodic basis Signed-off-by: marian-pritsak <[email protected]>
orchagent/pfcwdorch.cpp
Outdated
| @@ -22,6 +22,7 @@ | |||
| #define PFC_WD_POLL_TIMEOUT 5000 | |||
| #define SAI_PORT_STAT_PFC_PREFIX "SAI_PORT_STAT_PFC_" | |||
| #define PFC_WD_TC_MAX 8 | |||
| #define COUNTER_CHECK_POLL_TIMEOUT_SEC 5 | |||
There was a problem hiding this comment.
Is the polling interval controlled via cli or via config_db.json? If yes can you please explain how?
If not this needs fixing prior to committing.
There was a problem hiding this comment.
this is not controllable from the cli, this is only polling queue when the queue in pfc recovery state, in this case, we need to periodically poll it to get dropped packets. I do not see a generic need to make it controllable from cli.
There was a problem hiding this comment.
@lguohan All polling intervals need to be controlled via the cli and should provide the ability to be disabled.
orchagent/pfcwdorch.cpp
Outdated
| @@ -22,6 +22,7 @@ | |||
| #define PFC_WD_POLL_TIMEOUT 5000 | |||
| #define SAI_PORT_STAT_PFC_PREFIX "SAI_PORT_STAT_PFC_" | |||
| #define PFC_WD_TC_MAX 8 | |||
| #define COUNTER_CHECK_POLL_TIMEOUT_SEC 5 | |||
There was a problem hiding this comment.
Is the polling interval controlled via cli or via config_db.json? If yes can you please explain how?
If not this needs fixing prior to committing.
orchagent/pfcwdorch.cpp
Outdated
| @@ -22,6 +22,7 @@ | |||
| #define PFC_WD_POLL_TIMEOUT 5000 | |||
| #define SAI_PORT_STAT_PFC_PREFIX "SAI_PORT_STAT_PFC_" | |||
| #define PFC_WD_TC_MAX 8 | |||
| #define COUNTER_CHECK_POLL_TIMEOUT_SEC 5 | |||
There was a problem hiding this comment.
@lguohan All polling intervals need to be controlled via the cli and should provide the ability to be disabled.
|
retest this please |
|
@Nikos-Li , this is a internal feature and the user should not care about it and does not need to control that. |
…#473) '/sbin/reboot' randomly trigger a delay in shutdown path on some platforms, this delay caused the lags to timeout. Tested 'kexec -e' 1500+ times and never triggered the shutdown delay. Signed-off-by: Ying Xie <[email protected]>
…ch (sonic-net#473) * Add default bridge id for bridge port id of type PORT in virtual switch * Update recording Virtual switch now supports Queue TYPE, so tests must be updated * Address comments
For every action handler call commitCounters on a periodic basis
Signed-off-by: marian-pritsak [email protected]