[OpenAFS] fsck in metadevice using lvm in solaris 8

[Torbjorn Pettersson]

 My first concern when I read your letter weren't related to
wether the afs fileserver were interfering with the disk
structure, but after reading most of the other replies I figure
that that is probably the most likely problem. My first concern
however were that you had configured the metadevices
incorrectly, and had one of the partitions overlapping the
lvm database, or vice versa... Have you investigated what
happens if you don't have the fileserver running and perform a
similar set of operations?

Regards
Torbjörn Pettersson

Renato Arruda <rla9216@rit.edu> writes:

> Hi,
> 
> I am installing an openafs server on a sun fire v880. I have 4 hds in joined 
> using lvm (d30 -r c1t2d0s0 c1t3d0s0 c1t4d0s0 c1t5d0s0, from /etc/lvm/md.tab) 
> making a big metadevice /dev/md/rdsk/d30 with more than 100g.
> 
> I am using logging in that metadevice (from /etc/vfstab: /dev/md/dsk/d30 
> /dev/md/rdsk/d30        /vicepa afs     2       yes     logging)
> 
> This is what happens when i unmount and try to fsck it. (it always does that, 
> though if i try to mount it again it mounts fine, because the fs doesn't have 
> a flag saying that it's state is not okay)
> 
> # fsck /dev/md/dsk/d30
> ----Open AFS (R) openafs 1.2.6 fsck----
> ** /dev/md/dsk/d30
> ** Last Mounted on /vicepa
> ** Phase 1 - Check Blocks and Sizes
> ** Phase 2 - Check Pathnames
> ** Phase 3 - Check Connectivity
> ** Phase 4 - Check Reference Counts
> ** Phase 5 - Check Cyl groups
> FREE BLK COUNT(S) WRONG IN CYL GROUP (SUPERBLK)
> SALVAGE? [yn] y
> 
> BLK(S) MISSING IN BIT MAPS
> SALVAGE? [yn] y
> 
> SUMMARY INFORMATION BAD
> SALVAGE? [yn] y
> 
> 19 files, 37 used, 105001474 free, 4 AFS files (34 frags, 13125180 blocks, 
> 0.0% fragmentation)
> /dev/md/dsk/d30 FILE SYSTEM STATE SET TO OKAY
> 
> ***** FILE SYSTEM WAS MODIFIED *****
> /dev/md/dsk/d30: AFS file system partition was modified; forcing full salvage
> #
> 
> (result from a #df -k: /dev/md/dsk/d30      105001511   65581 103885915     1%    
> /vicepa)
> 
> I don't have many volumes in it currently, but i wonder in case of a system 
> crash if this will became an issue. I must admit it is the first time i have 
> ran with this sort of problem (the other afs servers i have managed had 80g 
> and less /vicepx partitions and were running linux (ext3), not solaris)...
> 
> Could it be that the openafs fsck program doesn't understand/can't handle 
> these lvm metadevices? the few files/volumes that are there are completely 
> unchanged by fsck.
> 
> Any insight would be _greatly_ appreciated.
> Thanks,
> 
> Renato
> -- 
> Renato Arruda <rla9216@rit.edu> http://www.rarruda.org/
> _______________________________________________
> OpenAFS-info mailing list
> OpenAFS-info@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-info
-- 
######################################################################
Torbjörn Pettersson               #  Email   tobbe@strul.nu
Vattugatan 5                      #  Web     www.strul.nu/~tobbe
S-111 52  Stockholm, Sweden       #
######################################################################