I'm doing some NDMP full and cumilative incrementals on some NAS data via NetBackup. I've noticed that the incs are taking the same amount of time as the fulls and are backing up roughly the same amount of data. The change rate on the NAS data isn't that great. I'm also doing the NDMP dumps from the SnapMirror destinations.
I was hoping to save some tapes by doing incs over nighlty fulls. Can anyone help please?
May not be relevant...
If the data structure had been changed by other Apps or activies, it could change file's attribute such time stemps, last access ,etc... it may think it is new set of data and consider all the files and new/updated one. Then it could take the same amount of time as initial copy...
Thanks for the reply.
I've found this links:
1. "ndmpd ignore_ctime" doesn't work from the console as the article suggests
2. the only reference I could fine to 'ignore_ctime' was in ndmpd options, i.e. 'ndmpd.ignore_ctime.enabled on'
I had this same problem at a previous job. You can't run more than 8 incrementals in a row without a full. After day 8, they start to work like a cumulative backup. This is inherint in the NDMP protocol to my knowledge.
We had a problem a bit like this with Networker.
When we checked the logs, it transpired that Networker was actually doing full backups even though we told it to do incrementals.
We came to the conclusion that the problem is that strictly speaking, NDMP doesn't support the concept of incremental backups (this was alluded to earlier in the thread). It supports either full, or multilevel differential backups.
To get the same effect as an incremental backup (we do a full backup each weekend and nightly incrementals), we had to manually create a schedule in Networker that specified the backup level numbers explicitly for each night to perform a differential backup against the previous night. This may well create additional overhead when you try to restore - given that we had to manually configure the differentials it's not clear to me if Networker is smart enough to know that it has to rebuild the full state from them - so if you end up doing this make sure you test it thoroughly.
My suggestion therefore is that you check your Netbackup configuration logs to see if it is simply doing full backups all the time.
It really is a shame that NDMP is treated rather like a poor second cousin by everyone involved. Given that EMC and others were involved in specifying NDMP, I was surprised that Networker, and apparently other major (and very expensive) enterprise backup tools, don't appear to have recipes for performing NDMP backups and restores along the lines of the industry standard, ie weekly full and daily incremental.
As always, YMMV ...
Thanks for taking the time to put a response together.
I'm really not liking NDMP becuase of this issue (which may be contributed to ignorance!), doing fulls every night is not practicable.
If I get further forard I'll repot back to the thread. In the meantime if anyone has any insight, I'd appreciate it.
We're having the same problem with the backup of the fileserver on our Netapp and I would never have suspected NDMP as the cause...
This forum is a great place!
I think the limitations are really on the client side. None of the client implementations are great. It would be nice if they would offer you a recipe to build incremental backups by configuring the thing appropriately, but that doesn't seem to be the case.
I'd really like to see Networker working with NDMP Snapmirror to tape.
The issue is down to the fact that I'm backing up SnapMirror destination volumes. Apparantly you can only do full backups, icremental dumps will never work .
Thanks for the input