Re: [Networker] Multiple drives for recovery?
2010-03-11 13:54:52
Tim Mooney wrote:
In regard to: [Networker] Multiple drives for recovery?, George
Sinclair...:
Anyone else seen this?
I run a browsable recover (CLI), requiring one full and three
incremental tapes. NW loads all tapes (6 drives in tape library) and
starts reading from each, rather than starting with the full and
working through one tape at a time, to the last incr?
This is on 7.5SP1 on RH Linux. Client is running an older 7.2.2 release.
It's funny that you should bring up something like this. I've been
meaning to post to the list about a major issue with parallelized
multi-volume recovers. I don't want to hijack your thread so I'll
post more details in a different thread.
Well, this all started because I ran a browsable recover (CLI recover
tool) last night from client A, pointing to an older 7.x server1. The
browsetime that I used was from back in January 2010. The recover
required 4 tapes (one full, three incrementals). It ran the conventional
way by loading each tape one at a time. About 7.5 GB of data was
recovered. Everything looked fine. The dates and times all looked
consistent. I then created an MD5 hash listing of all the recovered data
(file permissions, owner, group, checksum, file size, etc.).
Next, I moved the tapes over to the new 7.5SP1 server2's tape library. I
repeated the recovery, from the same client A, and this time it loaded
the tapes simultaneously and starts reading from all of them, and I'm
thinking: "What the heck!!!????". I don't notice any error messages or
overwrite prompts in the recover window, however. The recover completes
and indicates the same number of recovered files. **BUT**, when I
validated it against the MD5 hash listing from above, it reports a
number of directories with new time stamps - new as in the date of the
recover, not their original mod times. Only directories had new mod
times, but not all of them; some were fine. Otherwise, everything else
was identical. I thought it was odd that only the mod times for
*certain* directories were not preserved during the recovery, but
otherwise, all other files are perfect, as are a number of other
directories.
So, today, I repeated the recovery, but I first disabled all but one of
the drives to force NW to load tapes one at a time. Eureka! It worked
like a champ, and everything validated!
I should also note that before any of the recovers, I generated hash
listing of the CFI for client A on both servers, and they were identical
except for the directory structure under /nsr/index/db6. Otherwise, all
the files were the same and the same number of files. Moreover, nsrinfo
for the given date/time produces identical results for both servers.
Yes, we've seen NetWorker parallelize multi-volume recovers. Most of the
time it works pretty well. IIRC, this is something that was added in the
7.x series (earlier versions would always serialize volume access). It
used to be configurable by creating a file in /nsr/debug (do a substring
search of the mailing list archives for striped_recover for more info).
I find it hard to believe that NW can utilize multiple drives. How does
it merge and/or munge everything properly? What if you're instead
recovering from multiple fulls? How can it temporarily store all that
data as disk space could become jeopardized at some point. How does it
organize and/or re-conglomerate all that later?
Granted, this is a feature that I've always thought would be nice as it
would cut down restore times by many factors if it could recover in
parallel, but again, this raises my questions above. I wasn't aware that
this feature was ever developed and in use in later versions. I
generally watch the GUI when doing restores, and I've always seen tapes
loaded one by one on 7.2.x releases, but 7.5SP1 is all new to us.
We have, however, seen a few instances where recover apparently deadlocks
in the striped recovery code. This happened to us to a couple of times
under 7.2.x or 7.4.x, but we upgraded to 7.5.2 last week and the first big
recover we had to do triggered a deadlock in recovery. We've had a case
open with EMC about this issue since last Friday.
What do you mean by 'deadlocks'?
Do you think the parallel recovery would most likely explain the
weirdness that we see?
I was thinking to try upgrading the client software on client A, but I
doubt that has any effect over what the server decides to do on its end
in terms of loading those tapes. Also, it seems unlikely that the index
itself is somehow the culprit. Granted, it might know which tapes the
data is located on, but the server is still gonna handle the loading.
Moreover, the jukebox configuration and/or the tape library seems an
unlikely suspect as it just does what the server tells it.
If I have to disable drives during multi-tape restores that's gonna be a
real pain. sigh ...
George
Tim
--
George Sinclair
Voice: (301) 713-3284 x210
- The preceding message is personal and does not reflect any official or
unofficial position of the United States Department of Commerce -
- Any opinions expressed in this message are NOT those of the US Govt. -
To sign off this list, send email to listserv AT listserv.temple DOT edu and type
"signoff networker" in the body of the email. Please write to networker-request
AT listserv.temple DOT edu if you have any problems with this list. You can access the
archives at http://listserv.temple.edu/archives/networker.html or
via RSS at http://listserv.temple.edu/cgi-bin/wa?RSS&L=NETWORKER
|
|
|