Veritas-bu

[Veritas-bu] Long Running Vaults.

2000-07-17 13:50:39
Subject: [Veritas-bu] Long Running Vaults.
From: Brian Blake bblake AT veritas DOT com
Date: Mon, 17 Jul 2000 13:50:39 -0400
Lisa,

First, I would recommend opening a support call for vault. We have a dedicated
CSE to handle vault cases so we should be able to help you here.

Second, it really depends on the type of images being duplicated. If the image
is pretty broken up on the tape (check images on media report from NBU
reports), then it can take longer to duplicate since there may be extra
positioning involved. Also, are the images multiplexed at all? If so, are you
de-multiplexing the images during duplication (i.e. setting "mpxdup" to "1" in
the parm file)?

Have you changed the search criteria in the parm file? Have you increased the
number of days to search, or added any new classes or schedules?

If you want to check individual images, you can use the "All Log Entries"
report from NBU, and look for KBytes/sec. That will give you an estimate of
the performance during duplication. You can also turn on your bptm log and see
if there is any problem at that level.

Also, from what I have heard from our engineering group, bpduplicate uses the
TCP/IP protocol stack during dups... as opposed to using shared memory... so
there is a bit more overhead involved... hence a backup that was done locally
may take a little longer when you dup it.

Finally, you might want to try messing with the SIZE_DATA_BUFFERS and
NUMBER_DATA_BUFFERS parameters... these are created as files in
/usr/openv/netbackup/db/config. For more info, check out support.veritas.com,
go to the NetBackup UNIX section, and search the knowledge-base for "buffer".

Regards,

Brian

"Myers, Lisa" wrote:

> Hello!
>
> Lately, we have been experiencing long running vaults.  Seems like looking
> for a needle in a haystack.  Here are some of the things I have already
> checked:
> 1) recent changes to .param files for bpvault.all
> 2) tape density changes (all at 35- DLT7000 drives)
> 3) significant increase in data being duplicated (no)
> 4) other processes slowing down the system in general (i.e. performance
> monitoring programs- no)
>
> It appears that a Daily Vaulting is taking longer than twice what it was
> taking 2 weeks ago (were taking approximately 2-3 hours, now this week
> taking greater than 6 hours).
>
> Any trouble shooting tips that I am missing?  What have you used to track
> this kind of problem down?
>
> I am currently digging in the sar reports.  Any additional info. would be
> great!  Thanks in advance!
> _______________________________________________
> Veritas-bu maillist  -  Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu

--
Brian Blake
Professional Services Organization
VERITAS Software
brian.blake AT veritas DOT com





<Prev in Thread] Current Thread [Next in Thread>