My dremio runs in a machine with 32GB RAM and 8 CPU cores
this machine contains three dockers containers:
one with dremio (DREMIO_MAX_MEMORY_SIZE_MB = 16384)
one with mongodb (it sees 16GB of RAM)
one with a service in nodejs
It could be possible that when the planning time is high there are Full GC’s happening. I also see one another issue. You have distributed the total RAM of 32 GB between MongoDB and Dremio leaving nothing for the OS. Can you please GC logs when the query spends lot of time in planning?
On the Dremio container, do a “ps -ef | grep dremio”, look for “-Xloggc” in the output and you will see the path to server.gc. If you do not see that flag then it will be part of server.log
I have enabled this key.
But there are still queries that continue to take a long time (over 10 min) in “planning time”.
Is there anything relating to machine settings I can do?
Or can I limit the planning time and make the query cancel if it takes too long?