If this is a known problem, and the patch to resolve it exists, and it has been thoroughly tested (I assume this is done before customer support would provide the patch to a customer), then why isn't it posted in the s/w downloads or in the "Solution or Fix" (redundant?) item in the Bug Details table for Bug 31340? Why hasn't a Field Alert been posted? I would like to preemptively advise my customers who use NDMP compliant s/w to install this patch -- do they have to initiate a case with customer support to get it? Joe
Joe Luchtenberg Dataline, Inc. New number and email, please update: 757.858.0600 858.0606 fax joe.luchtenberg@data-line.com
-----Original Message----- From: Stephen Manley [SMTP:stephen@netapp.com] Sent: Wednesday, November 29, 2000 12:18 PM To: BrianH@dice.com Cc: toasters@mathworks.com Subject: Re: Unable to delete a snapshot
Hello, I have what could be a major problem soon. I have a snapshot that was created by an ndmpd session, that cannot be deleted. It looks like the backup must have failed, but the netapp thinks it is still running. Any ideas on how to delete the snapshot_for_backup.19?
dice-f740-1> snap list Volume vol0 working...............................................
%/used %/total date name
2% ( 2%) 1% ( 1%) Nov 29 08:07 hourly.0 7% ( 5%) 4% ( 3%) Nov 29 00:08 nightly.0 9% ( 5%) 6% ( 3%) Nov 28 20:29 oraclebackup 10% ( 5%) 7% ( 3%) Nov 28 20:04 hourly.1 14% ( 6%) 9% ( 4%) Nov 28 16:04 hourly.2 16% ( 6%) 12% ( 4%) Nov 28 12:07 hourly.3 21% ( 9%) 16% ( 6%) Nov 28 08:05 hourly.4 24% (10%) 19% ( 7%) Nov 28 00:06 nightly.1 26% (10%) 21% ( 7%) Nov 27 20:06 hourly.5 31% (11%) 26% ( 8%) Nov 23 21:04 snapshot_for_backup.19 (busy) dice-f740-1> snap delete snapshot_for_backup.19 Snapshot in use by dump or a CIFS share. dice-f740-1>
Hi everybody! Dr. Nick Riviera here!
I think Brian is actually hitting burt 31340:
%%% TITLE: ndmpd deadlock during aborts can cause a locked snapshot %%% DESCRIPTION: Aborting an ndmpd session could cause a locked snapshot to occur. The ndmpd session tries to abort the dump by repeatedly sending it a signal until the signal gets through. However, the dump is waiting for data being sent to the ndmpd session to drain first before processing the abort signal. This results in a deadlock situation where the dump thread does not terminate to release the snapshot. %%% WORKAROUND: Reboot the filer.
This is a filer bug. It is caused by using NDMP, but it is not the NDMP client's fault.
I believe if you contact customer support, you should be able to get a patch for this particular problem.
After getting the fix for 31340, no customer has had a problem with a snapshot being locked after the dump/NDMP has been terminated.
As for Bruce's request for a command to terminate dumps - we agree with you, and we're working on it.
So, I hope this helps.
Stephen Manley DAM and NDMP Mascot