Re: reiserfs/md1/failure/threads

[Francesco Pietra <frapietra@alice.it>]

I followed with interest high-level discussions on the filesystem problems for 
amd64 by Peter Yorke, Michael Marchand, Jo Shields, and Erik Mouw. However, 
at low level, I would be much obliged for suggestions about what to do with 
my system, where crash relateted to resiserfs 3.6 occurred while my data are 
intact. I suppose that scanning disks for possible defetcts, and recovering 
everything should be possible. How at best?
thank you
francesco

Added info below
On Tuesday 18 July 2006 10:09, Goswin von Brederlow wrote:
> Francesco Pietra <frapietra@alice.it> writes:
> > Because the matter is rather obscure to me, sent also CC to mpqc although
> > it seems an exclusive problem of the OS.
> >
> > OS: Debian amd64 etch
> > Hardw: Tyan S2895 K8WE; two 265 dual opteron; 8GB ram Kingston ECC;
> > raid1; temperature cpu low throughout.
> >
> > Process: mpqc 2.3.1 calc b3lyp geom optimization  MCsearch OO calculation
> > for a large molecule, max_iterations = 40, memory = 5GB, launched as
> >
> > $mpqc -messagegrp "<ShmMessageGrp>:(n=4)" filename.in | tee filename.out
> >
> > all 40 iterations were completed in a couple of days, with "Optimization
> > NOT converged".
> >
> > Restarted from the last minimum geometry, with memory = 7 GB, launched as
> >
> > $ mpqc -threadgrp "<PthreadThreadGrp>:(num_threads=4)" filename.in | tee
> > filename.out
> >
> > calculation hanged after ca 11 hours with warnings:
> >
> > ReiserFS: md1: warning: vs-5150: searchby key: invalid format found in
> > block 589839. Fsck?
> >
> > ReiserFS: md1: warning: vs-13070: reiserfs_read_locked_inode: i/o failure
> > occurred during tryong to find stat data of [7109 7110 0x0 SD]
> >
> > ReiserFS: warning: is_tree_node: node level 0 does not match to the
> > expected one 1.
> >
> > and several other similar warnings.
> >
> > Thanks for suggestions
> >
> > francesco
>
> If you checkd the disks, cables, controler and parity of the raid
Is any software command to partly fulfill your suggestions?
> then 
> the only suggestion I can make is don't use reiserfs.
When installing the OS I thought that  the most recent fs (reiserfs 3.6) were 
the most secure. Actually I have reiserfs on i386 (no raid) with no problem. 
Anyway, probably a warning against reiserfs on the installation disk or 
manual would avoid much troubles to users and advicers.

ADDED INFO: the system starts normally and I can read the out file from the 
mpqc calculation. I can recover the last geometry to restart the calculation.

Does this mean that the hardware is OK and it was only a failure of reiserfs 
3.6?

thanks again
francesco

Thanks a lot
francesco pietra
>
> MfG
>         Goswin