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
I am testing observer in a 12core machine. While there is no request, observer still cost 60% CPU of one core.
here are details:
I wrote a script to limit its cpu usage. seems that it works.
( this method requires that cgroup is enabled in your machine. create a cgroup named oceanbase)
PIDS=`top -d 10 -H -b -n1 | awk 'NR>6 { print $1,$12 }' | grep -E '(WeakReadSvr|LogStateDri|KVCacheWash|BRPC4|KVCacheRep|ObTimer|WeakReadService|ObLockWaitMgr|IODISK30|PartSerLargeCb3|CLGWR)' | awk '{print $1}'`
for i in $PIDS; do echo $i > /sys/fs/cgroup/cpu/oceanbase/tasks; done
echo 10000 > /sys/fs/cgroup/cpu/oceanbase/cpu.cfs_quota_us
# check
cat /sys/fs/cgroup/cpu/oceanbase/tasks
The text was updated successfully, but these errors were encountered:
MizuhaHimuraki
changed the title
observer cost too much CPU even if no user request at all
Observer cost too much CPU even if no user request at all
Jun 3, 2021
I wrote a script to limit its cpu usage. seems that it works.
( this method requires that
cgroup
is enabled in your machine. create a cgroup namedoceanbase
)The text was updated successfully, but these errors were encountered: