Kernel panic when rsyncing files to zfs pool. Root is on ZFS. Enters `mountroot'.

Hello.

The Problem
I am experiencing kernel panics and freezes on my FreeBSD 9.0 Release ZFS file server when I rsync or copy files to the pool where I store my critical data (i.e. not the root pool). This started after I installed a new pool of disks some weeks ago.

Root is on ZFS
I have root (i.e. FreeBSD installation) on a mirrored zfs pool on two old 80G IDE disks set to 4K sector size (http://www.aisecure.net/2012/01/16/rootzfs/) with swap as a "zfs partition" in that same pool.

History
My new data pool since some weeks ago (I call it `tank') is 8 * 2TB SATA HDDs set to 4K sector size in raidz2, which makes about 10T of storage.

What happens after
Because of the freezing sometimes I end up with mountroot shell because of unclean shutdowns. So I have reinstalled several times because I just cold not figure what to do on this.

Typical Chain of Events
I am thinking this is a memory problem. I have 10G DDR2 RAM in this box ... A likely scenario is this:
  • I observe in top(1) that the system with tank exported uses about 500M of ram.
  • When I import tank, the system chews another 500M of ram.
  • When I try to scp some 20G~50G large gzipped tarballs over to the root pool, I observe in top(1) that the system chews memory until it stops at about 7.5G. `zpool status' works ok, all is well.
  • When I try to rsync the archives over to tank, e.g. with parameter `--progress' on rsync, tank seems to freeze. I.e. rsync freezes, ls also freezes when I try to list files in tank, zpool status is not able to show info on either of the pools (it freezes, does not start to show anything), worst of all: I can not reboot, because the system is not able to execute the shutdown process fully, i.e. console just says something like `Shutdown is eminent. It was initiated by user rasmus', it never gets to spin down disks or anything of the sort. So I just hard reset the system ...
  • booting is going well sometimes, other times I get mountroot shell ... the times I am not, I try to scrub the pools, then the system runs into a reboot panic loop that can be ended by booting into LiveCD mode from a memstick and import the zroot then reboot, probably ending the scrub(s). If that does not work I just reinstall.

Other thoughts
Another possibillity would be that there are something wrong with the root disks, so that FreeBSD was never correctly installed, therefore it acts abnormal.

I suppose there are some options to /boot/loader.conf or the kernel config that can be added to make some sense out of this.

Does anyone care to help out?

/Rasmus

Edit: I made some headers to this post so it is easier to sort out important info + changed the title, since the old one do not make sense now.
 
Back
Top