On 1/24/22 10:25, Abhiram R N wrote:
> Making the log extack messages as ERR rather than DBG will help debugging
> when we dont know for sure where is the issue. If it is DBG these useful
> log extack messages might get lost due to excessive logs. Also these
> error messages will be printed only when we enable it.
> 
> netlink_socket(revalidator31)|DBG|Dropped 25 log messages in last 1 seconds
>  (most recently, 1 seconds ago) due to excessive rate
> 
> By checking for the log extack err_msg we are printing error message only
> when there is a valid log extack message present and other cases it will
> be still kept as DBG so that there is no spamming.

Hi, Abhiram.  Thanks for the patch!

The problem with this implementation, however, is that we're getting
a lot of errors in logs during the normal run.  For example, all the
system tests (make check-kernel) are failing for me due to unexpected
erorrs in the log.

OVS does a lot of probing to figure out what features are supported
by the underlying datapath and these may/will produce all sorts of
errors.  I also think that we will have more and more extended error
messages inside the kernel, so errors will be reported even more
frequently.  This makes testing harder and will trigger the monitoring
tools in production environments.  So, I'm not sure how the good
solution should look like.

For the rate limiting though, there is a solution.  You may disable
it for this particular module with the vlog/disable-rate-limit appctl.

Best regards, Ilya Maximets.
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

Reply via email to