Remove/Destroy VM

[kaltsi]

Hi,

I create and destroy lot of vm in a short time.
The destroy task usually very slow.
In the first 10 seconds the linked disk is removed.
I don't know what's happening in the background but the whole destroy task takes more then 1 minutes.

How can I found out why the destroy process takes long time?

 

[Stoiko Ivanov]

* hm - please post the complete task-log of a destroy operations
* else taking a look at the output of `ps auxwf` usually helps in finding out which process might be the reason for the long running time

I hope this helps!

 

[kaltsi]

Hi,

I could not analyze the whole process.
I found something.

This task took 1 minute and 2 seconds.

Every 1.0s: ps --forest -o pid,tty,stat,time,cmd -g 2572454                                                                                                                                                                                                                           vCNTDB-host-1: Mon Aug 26 16:30:40 2019

    PID TT       STAT     TIME CMD
2572454 ?        Ss   00:00:00 task UPID:vNTDB-host-1:002740A6:107773D4:5D63ECE6:qmdestroy:500:root@pam:
2572507 ?        Sl   00:00:13  \_ /usr/bin/rbd -p vNTDB-Storage -c /etc/pve/ceph.conf --auth_supported cephx -n client.admin --keyring /etc/pve/priv/ceph/vNTDB-Storage_ct.keyring ls -l --format json

root@vNTDB-host-1:~# time /usr/bin/rbd -p vNTDB-Storage -c /etc/pve/ceph.conf --auth_supported cephx -n client.admin --keyring /etc/pve/priv/ceph/vCNTDB-Storage_ct.keyring ls -l --format json

....

real    0m49.294s
user    0m16.430s
sys     0m1.418s
root@vNTDB-host-1:~#

It seems that it will take a long time to get this list.

The task log does not contain useful information.

why is it necessary to get whole list of chep after deletion?

 

[kaltsi]

Is it not enough to use the "/usr/bin/rbd ... ls" command? Without "-l".
It is faster.

real    0m0.054s
user    0m0.026s
sys     0m0.009s