dump and jails - system becomes unuseable then self heals

[Terry]

On 20/04/2010 23:31, Trever wrote:
> Hello,
>
> We have 4 servers that are running FBSD RELEASE 7.2 amd64.  The servers are each a jailer for 2 other jails (all basically identical configuration).  The jails were built per:
> http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/jails-application.html
>
> We run standard dump to back up the systems.  This is a procedure we use on a large number of other FreeBSD servers, none of which have the problem I'm about to describe.  (It is clear the problem is confined to these 4 systems that have jails).
>
> When dump runs on jailers (we don't run from within the jails), the system becomes unuseable eventually, then becomes available when the dump is finished (system freezes, unfreezes, etc. until the dump is done).  Apache spawns 256 children, etc.  We can't ssh.  Then we can.  We can't login on the console, then we can.  Etc.  So it goes until the dump has finished.  We have noticed that the system always responds to ping (simple ping checks from a load balancer do not detect the outage, whereas other health checks do detect).
>
> The relevant filesystem situation looks like this:
>
> df -h
> /dev/da0s1e             1.7T     22G    1.6T     1%    /data
> /data/jread/mroot       1.7T     22G    1.6T     1%    /data/jread/webs1j1
> /data/jread/mroot       1.7T     22G    1.6T     1%    /data/jread/webs1j2
> /data/jwrite/webs1j1    1.7T     22G    1.6T     1%    /data/jread/webs1j1/data
> /data/jwrite/webs1j2    1.7T     22G    1.6T     1%    /data/jread/webs1j2/data
>
> mount
> /dev/da0s1e on /data (ufs, local, soft-updates)
> /data/jread/mroot on /data/jread/webs1j1 (nullfs, local, read-only)
> /data/jread/mroot on /data/jread/webs1j2 (nullfs, local, read-only)
> /data/jwrite/webs1j1 on /data/jread/webs1j1/data (nullfs, local)
> /data/jwrite/webs1j2 on /data/jread/webs1j2/data (nullfs, local)
> devfs on /data/jread/webs1j1/dev (devfs, local)
> devfs on /data/jread/webs1j2/dev (devfs, local)
>
> cat /etc/fstab:
> /dev/da0s1e             /data           ufs     rw              2       2
> /data/jread/mroot       /data/jread/webs1j1                     nullfs  ro      0       0
> /data/jread/mroot       /data/jread/webs1j2                     nullfs  ro      0       0
> /data/jwrite/webs1j1    /data/jread/webs1j1/data        nullfs  rw      0       0
> /data/jwrite/webs1j2    /data/jread/webs1j2/data        nullfs  rw      0       0
>
>
>
> Anyone know what this could be?
>
> Thanks in advance,
>
> -T
>
>
>
>
>
> _______________________________________________
> freebsd-jail at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-jail
> To unsubscribe, send any mail to "freebsd-jail-unsubscribe at freebsd.org"
>    
I think if your dumping a live system with the -L switch it will lock 
things up while it takes a snapshot