Coda File System

rvm full ?

From: Ivan Popov <pin_at_medic.chalmers.se>
Date: Tue, 20 Jan 2004 03:32:59 -0500
Hello,

it looks like I am hitting rvm fragmentation or some other phenomenon that
makes further volume modifications fail (create or delete files).

It looks like I still can modify other volumes, but not the ones with most
objects in them.

I have one volume with 31176 files/dirs reported at server start.
Those are symlinks, dirs and very small files.
Now I cannot delete or create files in that volume anymore.
(the logs complain about no space in the volume log, trying to wrap around
and failing)

When I tried to copy the contents of the problematic volume to a new  one,
the first try seemed to succeed, but "tar" made some mistakes. Deleted the
new volume and retried with cpio -pVdm (outputting a dot per file).
Now the cpio log contains 34515 dots - it implies that it counts objects
in a different way than the server does...

At that point the copy was not yet finished, and both of my replicated
servers died (10 minutes one after the other), the first one logging:

Log:
----
 ...
05:55:58 AllocViaWrapAround: 0x2dfd.69ce has only single vnode on list

05:55:58 AllocViaWrapAround: 0x2e61.6b07 has only single vnode on list

05:55:58 AllocViaWrapAround: Gave up at 32 iterations

05:55:58 SpoolVMLogRecord - no space left in volume
05:55:58 - returns ENOSPC
----
Err:
----
 ...
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
Assertion failed: SpoolVMLogRecord(vlist, ov, volptr, &(VV->StoreId),
ResolveNULL_OP, 0) == 0, file "resclient.cc", line 165
EXITING! Bye!
----
                 And the second (having SrvErr 10 minutes newer)
Log:
----
 ...
06:04:08 AllocViaWrapAround: 0x3a5d.759e has only single vnode on list

06:04:08 AllocViaWrapAround: 0x3ad5.75dc has only single vnode on list

06:04:08 AllocViaWrapAround: 0x3b61.761f has only single vnode on list

06:04:08 AllocViaWrapAround: Reclaiming first log rec of 0xf.7649
----
Err:
----
 ...
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
No waiters, dropped incoming sftp packet
Assertion failed: vm_inuse->Value(*index) == 0, file "recov_vollog.cc",
line 618
EXITING! Bye!
----

Server and the libraries from cvs 2003-03-23.
Hmm, I do not recall that it would be too old with respect to rvm.

Would dump/restore rvm operation help against fragmentation?

Regards,
--
Ivan
Received on 2004-01-20 11:28:28