On 10/08/2013 04:59 AM, Indra Pramana wrote:
Dear Wido and all,
I performed some further tests last night:
(1) CPU utilization of the KVM host while RBD snapshot running is still
shooting up high even after I set global setting:
concurrent.snapshots.threshold.perhost to 2.
(2) Most of the concurrent snapshot processes will fail with either stuck
in "Creating" state, or "CreatedOnPrimary" error message.
Hmm, that is odd. It uses rados-java to call the RBD library to create
the snapshot and afterwards it copies it to Secondary Storage.
I'm leaving for the Ceph Days and the Build a Cloud Day afterwards in
London now, so I won't be able to look at this the coming 2 days.
(3) I also have adjusted some other related global settings such as
backup.snapshot.wait and job.expire.minutes, without any luck.
Any advise on the reason what causes the high CPU utilization is greatly
appreciated.
You might want to set the Agent log to debug and see if the RBD snapshot
was created, it should log that:
https://git-wip-us.apache.org/repos/asf?p=cloudstack.git;a=blob;f=plugins/hypervisors/kvm/src/com/cloud/hypervisor/kvm/storage/KVMStorageProcessor.java;h=1b883519073acc7514b66857e080a464714c4324;hb=4.2#l1091
"Attempting to create RBD snapshot"
If that succeeds the problem lies with backing up the snapshot to
Secondary Storage.
Wido
Looking forward to your reply, thank you.
Cheers.
On Mon, Oct 7, 2013 at 11:03 PM, Indra Pramana <in...@sg.or.id> wrote:
Dear all,
I also found out that when the RBD snapshot is being run, the CPU
utilisation on the KVM host will be shooting up very high, which might
explain why the host becomes disconnected.
top - 22:49:32 up 3 days, 19:31, 1 user, load average: 7.85, 4.97, 3.47
Tasks: 297 total, 3 running, 294 sleeping, 0 stopped, 0 zombie
Cpu(s): 4.5%us, 1.2%sy, 0.0%ni, 94.1%id, 0.1%wa, 0.0%hi, 0.0%si,
0.0%st
Mem: 264125244k total, 77203460k used, 186921784k free, 154888k buffers
Swap: 545788k total, 0k used, 545788k free, 60677092k cached
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
18161 root 20 0 3871m 31m 8444 S 101 0.0 301:58.09 kvm
2790 root 20 0 43.5g 1.6g 19m S 97 0.7 45:52.42 jsvc
24544 root 20 0 4583m 31m 8364 S 97 0.0 425:29.48 kvm
6537 root 20 0 0 0 0 R 71 0.0 0:17.49 kworker/3:2
22546 root 20 0 6143m 2.0g 8452 S 26 0.8 55:14.07 kvm
4219 root 20 0 7671m 4.0g 8524 S 6 1.6 106:12.26 kvm
5989 root 20 0 43.2g 1.6g 232 D 6 0.6 0:08.13 jsvc
5993 root 20 0 43.3g 1.6g 224 D 6 0.6 0:08.36 jsvc
Is it normal when snapshot is being run on the VM running on that host,
the host's CPU utilisation will be higher than usual? How can I limit the
CPU resources used by the snapshot?
Looking forward to your reply, thank you.
Cheers.
On Mon, Oct 7, 2013 at 7:18 PM, Indra Pramana <in...@sg.or.id> wrote:
Dear all,
I did some tests on snapshots since it's now supported for my Ceph RBD
primary storage in CloudStack 4.2. When I ran the snapshot for a particular
VM instance earlier, I noticed that this has caused the host (where the VM
is on) becomes disconnected.
Here's the excerpt from the agent.log:
http://pastebin.com/dxVV7stu
The management-server.log doesn't much showing anything other than
detecting that the host was down and HA is being activated:
http://pastebin.com/UeLiSm9K
Anyone can advise what is causing the problem? So far there is only one
user doing the snapshotting and it has caused issues to the host, I can't
imagine what if multiple users try to do snapshotting at the same time?
I read about snapshot job throttling which is described on the manual:
http://cloudstack.apache.org/docs/en-US/Apache_CloudStack/4.2.0/html/Admin_Guide/working-with-snapshots.html
But I am not too sure whether this will help to resolve the problem since
there is only one user trying to perform snapshot and we already encounter
the problem already.
Anyone can advise how I can troubleshoot further and find a solution to
the problem?
Looking forward to your reply, thank you.
Cheers.
