Operating System - HP-UX
1824994 Members
2164 Online
109678 Solutions
New Discussion юеВ

Re: NetBackup: Stats 25 - cannot connect on Socket

 
Alzhy
Honored Contributor

NetBackup: Stats 25 - cannot connect on Socket

Am implementing NBU 4.5 on HPUX11i environments. Master/Media server is on 1 server (N-Class 4 GB RAM, 4x650Mhz) connected to a 6 DLT7K Library presented as bridged devices over the SAN..

Except for sporadic Status 25 messages and network timeouts(my TOV is now 1t 1800, 100FDX infra btw..), the backups are running perfect. Are there any kernel/TCP settings that I am missing here?

Thanks!
Hakuna Matata.
3 REPLIES 3
Tim Sanko
Trusted Contributor

Re: NetBackup: Stats 25 - cannot connect on Socket

We found that we had similar responses on our L700 Using fibre channel to IBM LTOs. In fact we get a serious amount of EZ24 errors from our L700 --- No resonse from server.

As far as I know there really is little beyond bumping maxuprcs, that you could do that won't seriously muck up your server. Check performance and swapping. We have SUN v880s for our master server and it chokes and dies with a 25, or a 56-58 error when the CPU is at 100 % for all cpu's for long times.

Enable SAR and then you can see if you may be swapping (We were) and out of CPU resource. We ballanced our backup start times across our 2 Sunv880's and our 2 rp5470s and eliminated the 25's.

The catch is do you have the network bandwidth to handle the backup? We made a
dedicated subnet and that helped 50% of the problems go away. We went to gig e everywhere (fibre) and we saw all of the remaining 25s go away. Make sure your server has at least 2048 as maxuprcs ours is set at 4092.

Good Luck
Alzhy
Honored Contributor

Re: NetBackup: Stats 25 - cannot connect on Socket

It turned out I was locking maxuprc but the network was mostly the issue... there was a mismatch between my HBA and the switch.. Will be migrating the GigE in a few days.
Hakuna Matata.
Tim Sanko
Trusted Contributor

Re: NetBackup: Stats 25 - cannot connect on Socket

Gig E will not necessarily help the maxuprcs.
unless that the jobs are started and not completed because of throughput.

I would recommend looking at sar output, and/or measureware extract,exports.

I would look to see where the transactions fail (lanadmin to look for retries, dropped packets etc.)

I would also make the backup subnet segregated from production networks...

I just LOVE Veritas. (am I sarcastic or what)


Good Luck

Tim