You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This is very similar to #858, but I decided to open a new issue as the code involved is different in this case.
We have some metrics from production ~3 months ago that one ingester hit 400,000 goroutines. But little clear data.
To try to reproduce, I deliberately limited ingester's CPU and fired in 500 requests/sec from Avalanche.
So, when ingester can't keep up, many goroutines can be blocked on access to TSDB.
Just as in #858, it seems to me that once the number of goroutines goes beyond some limit we would be better off failing immediately than trying to carry on with the request.
The text was updated successfully, but these errors were encountered:
Just as in #858, it seems to me that once the number of goroutines goes beyond some limit we would be better off failing immediately than trying to carry on with the request.
In #3992 the support for an hard limit on the max number of inflight push requests.
This is very similar to #858, but I decided to open a new issue as the code involved is different in this case.
We have some metrics from production ~3 months ago that one ingester hit 400,000 goroutines. But little clear data.
To try to reproduce, I deliberately limited ingester's CPU and fired in 500 requests/sec from Avalanche.
Beginning of goroutine dump:
So, when ingester can't keep up, many goroutines can be blocked on access to TSDB.
Just as in #858, it seems to me that once the number of goroutines goes beyond some limit we would be better off failing immediately than trying to carry on with the request.
The text was updated successfully, but these errors were encountered: