04-30-2016 03:32 AM
I have a weird issue where half of the streams on an intelligent Oracle policy fail with status 58 but half will complete successfully.
If I set 12 streams 6 will run, 6 will fail. If I set 6 streams 3 will run 3 will fail. Anyone seen this kind of activity before?
Solved! Go to Solution.
05-03-2016 07:05 AM
nslookup is returning 2 names to the IP address. I am thinking it may well be something to do with that. Looking to get that resolved first.
04-30-2016 07:33 AM
Does the failing streams have the correct client name ?
Without more information, my best suggestion is to go thorugh the steps in posts below:
https://www.veritas.com/community/forums/general-database-backup-error-troubleshooting
If you havn't already. The database one hasn't been updated for OIP, but is still valid except for the manually running the script.
04-30-2016 11:03 AM
All connectivity tests passed. As some streams work I didn't think it would be connectivity.
6 streams start writing immediately after about 30 seconds 6 streams will immediately go to status 58. So it's not a timeout issue.
Failed streams have an unable to write progress log messages
04-30-2016 12:43 PM
04-30-2016 01:01 PM
That was my thinking as there is 50% success. Two interfaces on the client and they both have an seperate entry in the hosts file. Solaris 10 client btw.
04-30-2016 10:14 PM
Tbis isn't a RAC is it?
05-01-2016 12:13 AM
No. single server.
05-01-2016 05:20 AM
Do you have the required network interface set on the client? It has 2, but you don't want to give a choice on which one to use. Same thing from the master, 1 client, 1 name, 1 ip address.
05-01-2016 05:39 AM
Master server has only 1 interface.
On the client I have set the preferred network as:
PREFERRED_NETWORK = 0/0 ONLY "Client Interface IP"
but no joy. The host routing table indicates that backup traffic would go through the interface I want anyway.
05-03-2016 01:44 AM
I've successfully run flat file backups with multiple streams however.
05-03-2016 02:07 AM
You will need logs to troubleshoot.
I know that OIP use additional logs - best to look them up in NBU for Oracle manual.
dbclient (777 permissions) on the clientshould be a good start along with these server logs:
Master: bprd (NBU must be restarted when this folder is first created)
Media server: bpbrm
A work-around would be create old-fashioned script on the client and initiate it from Oracle policy.
If that works fine, go back to OIP and troubleshoot using logs...
05-03-2016 02:20 AM
Could you post the RMAN output?
Have you seen this? https://www.veritas.com/support/en_US/article.000024473
" During backup and restore operations, Oracle RMAN makes repeated calls to the System Backup Tape (SBT) API. Typically, there is minimal delay between the API calls. However, when using large data files, especially in bigfile tablespaces, the delays may be lengthy in some Oracle configurations. Lengthy delays make it difficult for NetBackup to know if Oracle is hung versus just delayed. If the delay is excessively long, then NetBackup will fail the job. "
Also sbttest is a very useful utility for troubleshooting API issues. http://www.dba-oracle.com/t_rman_111_sbttest.htm
05-03-2016 04:04 AM
Not an expert in RMAN but this is in the script that he policy creates:
INCREMENTAL LEVEL=0 PROXY
so I believe it is a proxy backup.
05-03-2016 04:51 AM
Have you tried an old fashioned script as per Marianne's suggestion?
05-03-2016 07:04 AM
Yeah, we currently run a workaround solution where an RMAN script dumps the backup to a share and this is then backed up as a flat file so I know that RMAN works. The intelligent policies are meant to replace that work around. This is the only server that is causing an issue at the moment. 7 other servers have successfully migrated and they are basically the same virtual Solaris server configuration.
05-03-2016 07:05 AM
nslookup is returning 2 names to the IP address. I am thinking it may well be something to do with that. Looking to get that resolved first.
05-03-2016 10:40 AM