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
{{ message }}
Repository navigation
[Bug] too many "send fragment timeout. backend id" error #8942
I had searched in the issues and found no similar issues.
Version
trunk
What's Wrong?
In some high concurrency scenarios, a large number of "send fragment timeout. backend id xxx" errors may appear in the fe.log. And all subsequent requests keep reporting this error and cannot be recovered.
What You Expected?
This is because in high load scenarios, the execution thread pool on the BE side is full, and new requests will enter the waiting queue of the thread pool.
However, the timeout period of the plan fragment request sent by FE to BE is only 5 seconds. The waiting time for requests entering the waiting queue may be very long, resulting in a large number of rpc errors send fragment timeout.
And subsequent requests will continue to enter the waiting queue, causing all subsequent requests to time out.
Activity
ChPi commentedon May 12, 2022
version 0.15。
modified
remote_fragment_exec_timeout_ms=600000, this problem still exists