Hi Mike,
You might look at your setting in the Defaults.conf and see if
adjusting the delay dismount parameter is set long enough... that
said, it would make sense to grab all the files in some cases,
however, that gets into concerns about blowing out the cache (stuff
that needs to be in cache Tier 1 is released to make room for
recycling copies). The best compromise then is probably to ensure
the tape has a chance to be reused before dismounting...
cheers!
Mike
On Dec 8, 2008, at 9:40 PM, mike cannon wrote:
> Thanks for all the responses, this fixed us up.
>
> sfind /qfs1 -damaged -exec undamage -c 2 {} \;
>
> I watched this draining operation and the re-archive of the data.
> It seems like it will stage one file off the tape to drain at a
> time and then archives that file before staging the next. In the
> mean time the stage tape ?idles? to long and gets ejected. Same
> thing on the re-archive, constantly ejecting a tape it will need
> shortly.
>
> I think we should be keeping the tapes in the drive longer and
> stream all the files off the tape while it is mounted before it
> starts to archive (or use a different drive in parallel).
>
>
>
> --
> mike cannon
> mikec at clemson.edu
> 864.650.2577 (cell)
> 864.656.3809 (office)
>
> computing & information technology
> 340 computer court
> anderson, sc 29625
>
>
> From: mike cannon <mikec at clemson.edu>
> Date: Mon, 8 Dec 2008 10:29:00 -0500
> To: <sam-qfs-discuss at opensolaris.org>, <sam-managers at list.ee.ethz.ch>
> Subject: recycler not draining
>
> I have a recycler issue. It does not finish draining tape in a
> timely fashion if at all. In the past I have had a similar problem
> but in those cases "showqueue -v" would show the files with an
> issues and explain what the issue was. I don't see the issue here.
> Sometimes it works like I expect, two recycler runs and tape has
> been labeled. Other times it goes on for months without completely
> draining...
>
> Here is an example of where once the tape was flagged to recycle it
> has not finished after 8 runs of the recycler.
>
> [root at samqfs1 /qfs1 #] cat /var/tmp/recycler.log | grep ti:100858 |
> grep archive
> Waiting for VSN ti:100858 to drain, it still has 456 active archive
> copies.
> Waiting for VSN ti:100858 to drain, it still has 17 active archive
> copies.
> Waiting for VSN ti:100858 to drain, it still has 17 active archive
> copies.
> Waiting for VSN ti:100858 to drain, it still has 17 active archive
> copies.
> Waiting for VSN ti:100858 to drain, it still has 17 active archive
> copies.
> Waiting for VSN ti:100858 to drain, it still has 17 active archive
> copies.
> Waiting for VSN ti:100858 to drain, it still has 17 active archive
> copies.
> Waiting for VSN ti:100858 to drain, it still has 17 active archive
> copies.
>
> sfind confirms the 17 active archive copies.
> [root at samqfs1 /qfs1 #] sfind . -vsn 100858
> ./cubs/cubsdb1/HPPRD_18Nov2008/oracle/hrapp.dbf.Z
> ./cubs/cubsdb1/HPPRD_18Nov2008/oracle/hrapp1.dbf.Z
> ./cubs/cubsdb1/HPPRD_18Nov2008/oracle/hrlarge.dbf.Z
> ./cubs/cubsdb1/HPPRD_18Nov2008/oracle/psimage.dbf.Z
> ./cubs/cubsdb1/HPPRD_18Nov2008/oracle/psindex.dbf.Z
> ./cubs/cubsdb1/HPPRD_18Nov2008/oracle/pstemp.dbf.Z
> ./cubs/cubsdb1/HPPRD_18Nov2008/oracle/psundo.Z
> ./cubs/cubsdb1/HPPRD_18Nov2008/oracle/pylarge.dbf.Z
> ./cubs/cubsdb1/HPPRD_20Nov2008/oracle/hphist.dbf.Z
> ./cubs/cubsdb1/HPPRD_20Nov2008/oracle/psindex.dbf.Z
> ./cubs/cubsdb1/HPPRD_20Nov2008/oracle/pstemp.dbf.Z
> ./cubs/cubsdb1/HPPRD_20Nov2008/oracle/psundo.Z
> ./cubs/cubsdb1/HPPRD_20Nov2008/oracle/pylarge.dbf.Z
> ./cubs/cubsdb4/PSDWH_19Nov2008/oracle/dw_data.dbf.Z
> ./cubs/cubsdb4/PSDWH_19Nov2008/oracle/temp.dbf.Z
> ./cubs/cubsdb4/PSDWH_19Nov2008/oracle/undo.dbf.Z
> ./poolenode/poolenode_clone/QPC009/1106744052.0
>
> [root at samqfs1 /qfs1 #] sls -D ./poolenode/poolenode_clone/
> QPC009/1106744052.0
> ./poolenode/poolenode_clone/QPC009/1106744052.0:
> mode: -rw------- links: 1 owner: root group: other
> length: 2060386304 admin id: 0 inode: 5109670.4
> damaged; offline;
> copy 1: ----D Nov 22 02:47 25f71.1 ti 100867
> copy 2: -r--D Nov 22 02:47 3fc58.1 ti 100858
> access: Nov 22 02:47 modification: Nov 22 01:53
> changed: Nov 22 01:53 attributes: Nov 22 01:52
> creation: Nov 22 01:52 residence: Nov 28 20:00
> [root at samqfs1 /qfs1 #] sls -D ./cubs/cubsdb1/HPPRD_18Nov2008/oracle/
> hrapp1.dbf.Z
> ./cubs/cubsdb1/HPPRD_18Nov2008/oracle/hrapp1.dbf.Z:
> mode: -rw-r--r-- links: 1 owner: oracle group: dba
> length: 128595115 admin id: 0 inode: 1221596.5
> damaged; offline;
> copy 1: ----D Nov 20 02:46 ffc7.1 ti 100846
> copy 2: -r--D Nov 20 02:46 58e9.1 ti 100858
> access: Nov 19 02:01 modification: Nov 18 00:15
> changed: Nov 20 02:25 attributes: Nov 18 01:42
> creation: Nov 18 01:42 residence: Dec 3 11:32
> [root at samqfs1 /qfs1 #] sls -D ./poolenode/poolenode_clone/
> QPC009/1106744052.0
> ./poolenode/poolenode_clone/QPC009/1106744052.0:
> mode: -rw------- links: 1 owner: root group: other
> length: 2060386304 admin id: 0 inode: 5109670.4
> damaged; offline;
> copy 1: ----D Nov 22 02:47 25f71.1 ti 100867
> copy 2: -r--D Nov 22 02:47 3fc58.1 ti 100858
> access: Nov 22 02:47 modification: Nov 22 01:53
> changed: Nov 22 01:53 attributes: Nov 22 01:52
> creation: Nov 22 01:52 residence: Nov 28 20:00
>
> [root at samqfs1 /qfs1 #] showqueue -v
> Filesystem qfs1:
> Scan list
> 0 2008-12-08 09:26:47 qfs1 12__ poolenode/
> poolenode_clone/QPC011
> 1 2008-12-08 09:52:21 qfs1 12__ itcback/itcback_clone/
> QFC007
> 2 2009-12-08 09:26:07 background ---- .
> Archive requests
> qfs1.qfs1.1.0 create 2008-12-08 09:26:14
> files:6 space: 33.753M flags:
> Start archive at 2008-12-08 10:26:14 | 500000 files | 32.0G
> bytes
> type:f ino:5088249 s:0/f:0 space: 32.500k time:1228744507
> priority:0
> poolenode/poolenode_clone/QPC011/4122251112.0
> type:f ino:5088248 s:0/f:0 space: 6.094M time:1228744510
> priority:0
> poolenode/poolenode_clone/QPC011/4105473898.0
> type:f ino:5088247 s:0/f:0 space: 32.500k time:1228744508
> priority:0
> poolenode/poolenode_clone/QPC011/4021587822.0
> type:f ino:5088245 s:0/f:0 space: 32.500k time:1228744509
> priority:0
> poolenode/poolenode_clone/QPC011/4055142254.0
> type:f ino:5088244 s:0/f:0 space: 27.532M time:1228744517
> priority:0
> poolenode/poolenode_clone/QPC011/3988033391.0
> type:f ino:5088243 s:0/f:0 space: 32.500k time:1228744513
> priority:0
> poolenode/poolenode_clone/QPC011/3954478960.0
>
> qfs1.qfs1.2.1 create 2008-12-08 09:26:14
> files:6 space: 33.753M flags:
> Start archive at 2008-12-08 10:26:14 | 500000 files | 32.0G
> bytes
> type:f ino:5088249 s:0/f:0 space: 32.500k time:1228744507
> priority:0
> poolenode/poolenode_clone/QPC011/4122251112.0
> type:f ino:5088248 s:0/f:0 space: 6.094M time:1228744510
> priority:0
> poolenode/poolenode_clone/QPC011/4105473898.0
> type:f ino:5088247 s:0/f:0 space: 32.500k time:1228744508
> priority:0
> poolenode/poolenode_clone/QPC011/4021587822.0
> type:f ino:5088245 s:0/f:0 space: 32.500k time:1228744509
> priority:0
> poolenode/poolenode_clone/QPC011/4055142254.0
> type:f ino:5088244 s:0/f:0 space: 27.532M time:1228744517
> priority:0
> poolenode/poolenode_clone/QPC011/3988033391.0
> type:f ino:5088243 s:0/f:0 space: 32.500k time:1228744513
> priority:0
> poolenode/poolenode_clone/QPC011/3954478960.0
>
> --
> mike cannon
> mikec at clemson.edu <mailto:mikec at clemson.edu>
> 864.650.2577 (cell)
> 864.656.3809 (office)
>
> computing & information technology
> 340 computer court
> anderson, sc 29625
>
>
> _______________________________________________
> sam-qfs-discuss mailing list
> sam-qfs-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/sam-qfs-discuss
Mike Selway
Sr. Consulting Engineer
301-332-4116
Solutions Factory Team
Sun Microsystems
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://mail.opensolaris.org/pipermail/sam-qfs-discuss/attachments/20081208/a89edb9a/attachment.html>
