GitHub user weizhouapache added a comment to the discussion: Degraded 
cloudstack agent

> I tested libvirt (10.0.0 and 10.6.0) on two Ubuntu 24.04 nodes, and monitored 
> some metrics every 10 seconds by commands
> 
> ```
> pid=$(pgrep -f /usr/bin/java)
> echo "CPU: $(ps -p $pid -o %cpu --no-headers)%"
> echo "MEM: $(ps -p $pid -o %mem --no-headers)%"
> echo "FD: $(ls /proc/$pid/fd 2>/dev/null | wc -l)"
> echo "Threads: $(ps -L -p $pid -o tid --no-headers 2>/dev/null | wc -l)"
> echo "Conn: $(ss -tanp | grep "pid=$pid" | grep ESTAB | wc -l)"
> ```
> 
> My observations are
> 
> * There are very slight changes with memory, file descriptors, threads and 
> connections of the java process, I think we can ignore.
> * The CPU usage of the java process increased gradually and slowly with 
> libvirt 10.0.0
> * The CPU usage of the java process decreased gradually and became stable 
> after some hours with libvirt 10.6.0
> 
> Hostname      libvirt version original CPU usage      new CPU usage
> kvm1  10.0.0-2ubuntu8.9       0.9%    2.1% (after 12 hours), 2.8% (after 16 
> hours)
> kvm2  10.0.0-2ubuntu8.9       5.3%    5.9% (after 1 hour), 6.5% (after 2 
> hours)
> kvm2  10.6.0-1ubuntu3 4.9%    1.3% (after 1 hour), 0.6% (after 3 hours and 6 
> hours)
> kvm2  10.6.0-1ubuntu3.3       5.2%    1.2% (after 1 hour), 0.8% (after 2 
> hours)
> note:
> 
> * no workloads inside VMs
> * no VMs actions
> * original CPU usage is fetched 5 mins after restarting agent
> 
> I would suggest some of you test `libvirt 10.6.0-1ubuntu3.3` or `libvirt 
> 10.6.0-1ubuntu3` on a non-production environment. I have uploaded the .deb 
> packages to https://github.com/weizhouapache/ubuntu-libvirt-packages

Quick update:
- on kvm1 (with libvirt 10.0.0-2ubuntu8.9), CPU usage has increased to 7.5% 
after around 24 hours
- on kvm2 (with libvirt 10.6.0-1ubuntu3.3), CPU usage is still 0.6% after 
around 12 hours


GitHub link: 
https://github.com/apache/cloudstack/discussions/12450#discussioncomment-15515842

----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]

Reply via email to