From: Herbert Poetzl (herbert_at_13thfloor.at)
Date: Mon 25 Apr 2005 - 16:53:46 BST
On Mon, Apr 25, 2005 at 10:27:07AM +0200, alexx wrote:
> To begin : sorry for my poor english :(
> I've a recurent problem with vserver 0.30.204 on a bi-XeonP4 (use in
> 32bits, and
> with hyper threading).
> Kernel is 126.96.36.199 with vs1.9.5
> The host is debian, but i've compiled vserver-util alpha myself.
> I run 2 vservers on since approximativly 1 mounth.
> Vserver and host server run perfectly .. but last night, and this night, 1
> of the
> 2 CPU go into 100 % 'wa' (i don't find what that mean in 'top') and the
'wa' is I/O wait ... i.e. the cpu spends most of the
time waiting for some I/O transaction to complete ...
> other look
> like ok with 95% idle.
> The load average grow to 100 this night and 300 last night (normal load:
> max 4) !
> At this step, i have only one vserver down (stop responding) on the 2.
> The strange thing was it's not the same vserver last night, and this night
> When the "bug" arrived, i can't do "vserver .... enter" or "vps auxw", ...
> I can't enter into the directory of the vserver, ... the term freeze after
> [Enter] to the '$ cd /vservers/serv1' command
> just vtop work, but not display any strange process :( just an enormous
> "vserver .... stop" freeze too after a
> All VServer are on it's own LVM partition in ReiserFS.
> each partition is snapshoted a multiple of time.
> Last days modification is the add of a cron script to make new snapshot
> each night
> without delete the old (i want to keep 5 snapshots each time).
> I umount old snapshot, and mount the new snapshot with :
> "mount -o ro,noexec,nodev,noatime ..."
> The snapshot cron is at 0:30
> My backup server take data into the mounted snapshot at 1:00
> And the vserver crash no longtime after.
well, I would suspect that combination of LVM, reiser and
dm snapshooting to cause unusual hight I/O traffic ...
> I stop this for now (just one snapshot at a time per partition).
> Any other suggestion for the origin of the problem ?
check the kernel logs maybe they give some hints ...
Vserver mailing list