Unable to kill a vmtar process

[rengiared]

hi,

i have the problem that i can't kill a hung up vmtar process from a backup-procedure.
The problem starts with a strange behavior after a few minutes of a "big" machine started to backup.
For example, the backup of a smaller one with a 8GB tar needs about 16 minutes.
When i start to backup the "big" one with about 42GB, the first 15GB are loaded to the NAS very quickly, but then it gets unbelievable slowly.
Now i tried to kill the vmtar process cause i wanted to try something else but the process doesn't kill.
with "ps -C vmtar" i search the process-id, then i try to kill it with "kill 1234" but nothing happens
as a strange sideeffect every machine on the host gets incredibly slow and uses just the half of his assigned RAM.

i have the setting "size: 2048" in the vzdump.conf.
the host is a DELL R710 with 2 Xeon E5520, 24GB RAM and 8 146GB SAS Drives in a Raid 5 with 9 guests running on it.

in the end i have 2 questions. What else can i try to kill the process and why the backup gets so slow after a while.
the third would be how the strange sideeffect could be, but if i get 1 and 2 to work i hope 3 dissolves in air.

thanks in advance

 

[dietmar]

What else can i try to kill the process and why the backup gets so slow after a while.

If nothing helps, try:

# kill -9 1234

What is the output of

# pveperf

Please run when the host is idle.

 

[rengiared]

Hello,

a few minutes after my post the process got killed, next time i will try it immediately with kill -9. thanks

pveperf results are: (as i have no real idle situation i just made it now)
CPU BOGOMIPS: 72349.68
REGEX/SECOND: 814612
HD SIZE: 94.49 GB (/dev/mapper/pve-root)
BUFFERED READS: 94.17 MB/sec
AVERAGE SEEK TIME: 6.89 ms
FSYNCS/SECOND: 490.27
DNS EXT: 58.92 ms
DNS INT: 2004.64 ms

the high dns int value is a little bit strange