RE: [ogfs-users]mount freeze on 3rd machine

Hi Arnaud,

There is a known issue with the current support for OpenDLM.  It will
likely cause problems if you mount more than one OpenGFS filesystem on a
given node (is this what you mean by "partition"?).  We need to get some
static variables, used for recovery aspects, placed into an instance
structure (one instance for each mounted filesystem).  I'm not *sure*
that this is the root cause of your problem, but it might be.

Also, what does "thist" mean?  (Sorry for my ignorance).

BTW, realize that you're working in untested territory!!  Stan just
recently wrote the recovery support for using OpenDLM, and we've never
tried it with more than two nodes.  You are on the "bleeding edge", and
we thank you for it!  ;-)

For more output from the OpenGFS OpenDLM lock module, you can load it
with:

# modprobe opendlm debug=3D1

That will generate a lot of output, but it may be helpful.  I'm guessing
that the problem is in the locking, not the main filesystem code.

If you're eager to get going, and could tolerate a
single-point-of-failure, you could also try the memexp locking
(HOWTO-nopool).  You can switch back and forth between memexp and
OpenDLM, if you'd like (requires unmount/mount).  There is no on-disk
locking data (beyond the name of the default locking protocol, stored in
the superblock, which you can override in the mount command).

-- Ben --=20

> -----Original Message-----
> From: [email protected]=20
> [mailto:[email protected]] On Behalf=20
> Of Arnaud Gauthier
> Sent: Thursday, June 24, 2004 5:17 AM
> To: [email protected]
> Subject: [ogfs-users]mount freeze on 3rd machine
>=20
> Hello all,
>=20
> Thanks for helping me with CCM, it works fine now: I have a=20
> really fast
> CLM_ST_RUN status on all 3 machines.
>=20
> Here is my OpenGFS related problem (was also the reason I tried CCM)
> When I mount my 2 OGFS partitions on the thist machine, before it was
> usually freezing one of the previously mounted machines, and now the
> mount freeze one the mounting machine, but nothing else freeze (much
> better, before I was loosing a machine :-))
>=20
> Here is an extract of my messages log:
>=20
> [HSM:hsm_process_event] enter( me =3D 0xf9d1f2c0, msg =3D 0xf5895fe0 )
> [HSM:hsm_process_event_private] DLM recovery state machine (71 ->
> RC_DIR_INIT)
> [HSM:hsm_execute_transition] DLM recovery state machine (CLM_ST_TOP ->
> CLM_ST_RUN)
> [HSM:hsm_process_event] exit
> opendlm_start, debug =3D 0
> opengfs-0.3.0 built on Jun 23 2004
> sizeof(ogfs_meta_header_t) =3D 24
> sizeof(ogfs_sb_t) =3D 352
> sizeof(ogfs_jindex_t) =3D 80
> sizeof(ogfs_rindex_t) =3D 96
> sizeof(ogfs_rgrp_t) =3D 128
> sizeof(ogfs_indirect_t) =3D 88
> sizeof(ogfs_dinode_t) =3D 232
> sizeof(ogfs_dirent_t) =3D 40
> sizeof(ogfs_leaf_t) =3D 104
> sizeof(ogfs_log_header_t) =3D 128
> sizeof(ogfs_log_descriptor_t) =3D 104
> sizeof(ogfs_block_tag_t) =3D 16
> =20
> sizeof(ogfs_bitmap_t) =3D 12
> sizeof(ogfs_rgrpd_t) =3D 296
> sizeof(ogfs_glock_t) =3D 184
> sizeof(ogfs_bufdata_t) =3D 72
> sizeof(ogfs_inode_t) =3D 444
> sizeof(ogfs_trans_t) =3D 96
> sizeof(ogfs_sbd_t) =3D 34480
> sizeof(ogfs_dcached_t) =3D 8
> mtrr: type mismatch for f9000000,800000 old: uncachable new:
> write-combining
> mtrr: type mismatch for f9000000,800000 old: uncachable new:
> write-combining
> ar_lockproto =3D ""
> ar_locktable =3D ""
> ar_hostdata =3D ""
> ar_dir_alloc_method =3D 612324
> ar_ignore_local_fs =3D FALSE
> ar_localflocks =3D FALSE
> ar_localcaching =3D FALSE
> ar_buf_readahead =3D 8
> ar_atime_quantum =3D 60
> ar_upgrade =3D 0
> harness: found protocol opendlm
>=20
> journals are 512 MB big (I have 32 journals), one partition is 250GB
> big, the other 500GB... Access from the 2 other machines remains fine.
>=20
> Is there a way to receive more output from ogfs ?
>=20
> Regards,
> Arnaud
> --=20
> Arnaud Gauthier <[email protected]>
> Realmedia
>=20
>=20
>=20
>=20
> -------------------------------------------------------
> This SF.Net email sponsored by Black Hat Briefings & Training.
> Attend Black Hat Briefings & Training, Las Vegas July 24-29 -=20
> digital self defense, top technical experts, no vendor pitches,=20
> unmatched networking opportunities. Visit www.blackhat.com
> _______________________________________________
> Opengfs-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/opengfs-users
>=20
>=20


-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com