[PVE-User] Crash after Upgrade PVE2.3 / Cron backup crashes with 2.6.32-19 but not with 2.6.32-17

[Martin Schuchmann]

Hi Dietmar,

I had increased the size in the past to 4096 since there had been 
problems with canceled backups as discussed in the forum also, but they 
never crashed the whole system.

# vzdump default settings

#tmpdir: DIR
#dumpdir: DIR
#storage: STORAGE_ID
#mode: snapshot|suspend|stop
#bwlimit: KBPS
#ionice: PRI
#lockwait: MINUTES
#stopwait: MINUTES
size: 4096
maxfiles: 3
#script: FILENAME
#exclude-path: PATHLIST

And this setting works perfectly with  2.6.32-17, and also with 
2.6.32-19 with manual backups from the webinterface, but not with a 
scheduled one via cron with 2.6.32-19 on *this *node. On the two other 
nodes there had been no problems same time with scheduled backups (with 
much bigger VMs and CTs). The backup crashed even on a small, 
non-channging (unused owncloud file server)  CT with only 6GB HDD.

Please take a look on the lvdisplay output during a small, non system 
destroying crash:

    Allocated to snapshot 60,92%

So there should be enough space left?


I recognized the following difference between non-failing/failing:


After a crash the LV showed:

    LV Size                4,00 GiB
    Current LE             1024
    Segments               1
    Allocation             inherit

During a non-failing backup it was

    LV Size                2,55 TiB
    Current LE             669651
    COW-table size         4,00 GiB
    COW-table LE           1024
    Allocated to snapshot  60,92%
    Snapshot chunk size    4,00 KiB
    Segments               1

The 2,55 TiB in the working one correspond to the size of the data-LV:

   LV Path                /dev/promo3/data
   LV Name                data
   LV Status              available
   # open                 1
   LV Size                2,55 TiB
   Current LE             669651
   Segments               1
   Allocation             inherit


I recognized during the crashingbackups, not to be able to do an 'ls 
/mnt/pve/'  - this ends with a hung and no output during the backup. 
While a non-failing backup is running, there occurs no problem with 
that. 'lvscan' shows same behaviour.

Funny seems also this difference between lvdisplay and lvscan during a 
non-failing backup:

lvdisplay:

  --- Logical volume ---
   LV Path                /dev/promo3/vzsnap-promo3-0
  ...
   LV Status              available
   # open                 1
   LV Size *2,55 TiB*
   Current LE             669651
...

lvscan:
...
   ACTIVE   Original '/dev/promo3/data' [2,55 TiB] inherit
   ACTIVE   Snapshot '/dev/promo3/vzsnap-promo3-0' [*4,00 GiB*] inherit


I could try again the new kernel with a higher size in vzcron.conf - but 
it seems to me not to be the cause of a whole system crash. Even if the 
size-parameter may be to small - in my opinion there should be no chance 
for crashing the whole node with that?

many regards,

Martin



Dietmar Maurer <dietmar at proxmox.com> schrieb am 22.03.2013 06:36:
>> snapshot: Unable to allocate exception.
> You run out of snapshot space! You should increase that (see 'man vzdump' - parameter 'size').

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.proxmox.com/pipermail/pve-user/attachments/20130322/06b460b4/attachment.htm>