Hi,
we had a for the most part well working NBU system with a NBU 6.5.3 master/media server on Linux, a NBU 6.5.3 media server on Linux and two NBU PureDisk 6.5.1 systems. Usually, all the NBU clients on this system are backing up their data through the PDDO on the media server to the first PureDisk device in the first stage. On the second stage, there is a duplication running to a second PureDisk device (directly between the two SPA ). We are using storage lifecycle policies (SLP) for that.
Some days ago, we updated the whole system and are now on NBU 6.5.4, PDDO 6.6 and PD 6.6.
After the update, all the jobs with smaller than 50GB in size were working, but we saw bad performance, bad dedupe and bad cache hits, not only on the first jobs after the update. It is still happening.
All jobs with 50GB in size or greater failed, but this could be solved meanwhile by changing maximum fragment size on the PureDisk STU to a value that is lower than the value in pd.conf (MAX_IMG_MBSIZE) on the PDDO!MAX_IMG_MBSIZE is remarked by default, which means, it's on 50.000 (MBytes).
So, we set the maximum fragment size on the PureDisk STU to 48 GBytes = 49.152 MBytes.
The current problem is:
All the Backup jobs are still taking around 40% longer and the dedupe rate has gone down from > 90% to around 30...70%.
Because of the bad dedupe, too much backup storage on the PureDisk is used.
Anyone here, who had similar problems and found a solution?
Matthias
