GetMetrics took 37 ms, out of which callback took 37 ms

[SiebelsTim]

From time to time, we have a lot of output in the following form on stderr

GetMetrics took 37 ms, out of which callback took 37 ms

Is this something to we need to investigate? If not, would you be open to change the output to stdout rather than stderr?

[kostasrim]

Hi @SiebelsTim, logging goes on stderr and it's the usual convention.

Is this something to we need to investigate?

Not really, it's just a warning that info command is taking longer than expected (could be because of high load, etc). As long as your datastore is responsive there is no action here.

If not, would you be open to change the output to stdout rather than stderr?

All of our logs go to stderr. You can use --alsologtostdout which will also redirect to stdout but it will do it for all logging

[SiebelsTim]

Oh, I didn't realize that all logging goes on stderr. I will handle this in our monitoring then. Thanks!

Is that a CPU-bound task or is it affected by network? So that I can look into it under what circumstances that's happening more often.

[kostasrim]

mostly CPU-bound with a few exceptions (like replica takeoever) which will soon be fixed (there is an open PR) :)