My leader FE node is completely unresponsive, what do I do? I have removed all followers, I have done the bdbje_reset_election_group = true reset, it does start up initially, then CPU usage EXPLODES (I have 32 cores and 128gb ram on this machine)... This is where it freezes up and just stays like this indefinitely. I have removed ALL materialized views. I have removed ALL routine loads. There should be nobody using this leader FE node, yet after 1-2 minutes of starting it, it just exploes in CPU usage and becomes unresponsive to the point, where I have to power off the server in my cloud console. What is going on? We cannot lose our data yet again.

monitor the gc of the java thread, if the -Xmx is enough for the FE to complete is meta data loading and editlog replay.

Yeah it was definitely a GC issue! It was clear when I inspected CPU usage closer. Then it was just a matter of tweaking the JVM settings and the problem was gone. We are now fully operational again and we have lost no data 😄

good

@Sebastian Hindhede just curious, can you share the command you use to check the CPU usage that shows it's caused by the GC?

Yeah first I got the PID of the starrocks fe process

sudo ps aux | grep -i starrocks | grep -v grep

And then when I had the pid I ran

top -H -p <pid>

thanks! @Sebastian Hindhede