I just returned from a two day DR test through while successful it had its challenges on the Netbackup front. I was wondering if anyone else had experienced these kinds of challenges at a DR test.
Building the Netbackup (7.5.0.7) masters (one each on Windows, HPUX, and AIX) went flawlessly. But on the AIX platform we kept having the Java Client become very slow or unresponsive. I often had to kill the client and restart it. We did not have this problem on the HPUX master.
To the best of out knowledge the AIX backup master was not underpowered.
And on both the HPUX and AIX backup masters we submitted restores from the standard backups where we need to select perhaps a dozen or so file systems that are not in the Ignite/mksyysb images. We'd submit the backup and the job would sit in queued state for a long time. In one case it sat for six hours and then we cancelled it. On the HPUX platform the bprd process was running at 100% of one CPU. We did not see this on the AIX platform.
We ended up submitting the restores one file system at a time and then the job would go active immediately. But of course this was a terribly inefficient restore; we ended up spinning through the same tape multiples times and it was a very slow recovery.
In some of these cases we know that there would have been tens of thousands of files within the file systems being selected. But given I was selecting at a high level (the file system) - and you'd see only a dozen or some entries under the "File List" when looking at the job in the Activity Monitor GUI - does it matter how many files exist within those file systems? Does that slow down the startup of the job?
Given that Netbackup is so heaviliy dependent on DNS lookups working both forwards and backwards we suspect we might have had a problem there that slowed down Netbackup compared to how it runs here at home. Does anyone have any experience on what they look for to see if that is working as best as it can?
Any ideas/discussion we could have would be appreciated.