The advice you've got is correct - with the number of media servers that
you have, all drives should be zoned to all media servers. You should use
the Device Configuration Wizard on the Master Server to set this up - do
not use tpconfig on the media servers unless you are very, very sure what
you are doing. The wizard does a lot of checking that all definitions are
consistent. If you do it using tpconfig and anything on any media server
is 'a little different' it will not work properly. Don't get me wrong, we
script CLI stuff all the time, but in this case, use the wizard.
As you are using 6.5 I understand that you can actually share the drives
currently dedicated to the NetApp, but I've not been there as our NetApp
NDMP backup is not on 6.x yet. You do need to be a little careful as it
can mean that non-NDMP backups can be sent to the NDMP drives and
vice-versa - there was a touch file to disable this at 5.1, not sure if it
is still used at 6.5.
If your system grows don't let too many media servers share a drive, or it
can get out of hand. We've got some drives with silly numbers of servers
having access - there are only 24 hours in a day and if a drive is in use,
it is busy - adding more servers to a drive doesn't add resources! It can
make big problems if your FC is not correctly configured to limit the
impact of a server reboot. Do it wrong and a reboot of any server sharing
the drive will reset the drive. If your are using the default settings
they may be the wrong settings, as vendors often default to settings for
disk, which don't suit tape at all (e.g. queue depth). If you have
multipath disk, the vendor kit (Powerpath, Securepath etc) may set very
unsuitable settings as timeouts for multipath disk failover are very, very
wrong for tape.
2 LTO3 drives seems low for quad core servers. It used to be something
like 1 GHz CPU per 1GbE interface plus 1.5GHZ CPU per tape drive. I'd
expect that kind of server to cope with 2 x LTO4 or maybe 4 x LTO3 -
depends on compressibility of data, whether you are backing up D2D2T and
many other things. As always, look at the bpbpkar logs on the clients,
and the bptm logs on the media servers, and look at the wait & delay
times. This rather ugly scrap of shell script I use for testing, it looks
at the bptm log and will show if you are waiting for empty buffers (so the
tape drive is slowing you down) or the full buffers (the clients are not
sending data fast enough):
#!/usr/bin/ksh
#*+
#* Script to read a bptm log and print out wait & delay figures.
#* Takes one parameter, the log file name with path is not in the current
directory.
#* -
awk ' \
/waited for empty/ \
{ \
if ( $20 == "bytes" )
{ i_data=$19/1048576 }
else
{ i_data=$19/1024 };
printf ("%s\t%s\t%d\ttimes\t%s\t%.2f\tminutes\t%s\t%8.2f Mbytes\n", \
$1,$9,$13,$15,$16/3000,$18,i_data) \
}' $1
awk '/waited for full/ { OFS="\t"; printf
("%s\t%s\t%d\ttimes\t%s\t%.2f\tminutes\n",$1,$5,$9,$11,$12/2000) }' $1
Example output:
04:05:40.245 waited 6943 times delayed 12.23 minutes
Looking at it I've assumed that you have not adjusted the wait or delay
factors. If you look at the delay times and compare them with the total
elapsed time for the job you can see where time is wasted. In your case
you want to know if the media server is coping with the load, so I'd guess
you want to aim for 'waited for empty' being non-zero as that would show
that your media server was keeping up with the tape drives.
Now if your media servers are Windows...mmmm.
William D L Brown
veritas-bu-bounces AT mailman.eng.auburn DOT edu wrote on 30/06/2009 21:54:38:
> I have 1 windows master running Netbackup 6.5 with 3 media servers.
> I have Storagetek/Sun SL8500 with 8 LTO3 tape drives (2 drives are
> dedicated via NDMP to a Netapp 3050). All drives are connected via
> FC to brocade etc. we have the Shared storage option but I don?t
> think it?s working properly. How can I verify that its working?
>
> I believe the engineers that configured the zoning on the switch
> dedicated 2 drives per server and dedicate 2 to the Netapp. Did it
> makes sense to purchase SSO?
>
>
> I was told you should only have 2 LTO3 drives per servers because of
> ?shoe shining? and the servers ability to handle more than 2 drives
> (Dell 2950?s quad cores with 8gig of ram). I am wondering if the
> switch should be rezoned to make use of SSO or can the zoning have
> an impact on how SSO should be configured? How can these 2 play
together?
>
> As you can see I am all confused and would love for someone to
> explain this in simple terms.
>
> thanks _______________________________________________
> Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
> http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
-----------------------------------------------------------
This e-mail was sent by GlaxoSmithKline Services Unlimited
(registered in England and Wales No. 1047315), which is a
member of the GlaxoSmithKline group of companies. The
registered address of GlaxoSmithKline Services Unlimited
is 980 Great West Road, Brentford, Middlesex TW8 9GS.
-----------------------------------------------------------
_______________________________________________
Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
|