Is there a way to copy the cache files on the NDMP node for a client that I have configured to another NDMP node?
What I have going on is we have our file server at our HQ running on our vnx. The night previous I failed the file server over to our DR site while we had some code upgrades done on our Prod vnx. In turn now I need to get backups working on the file server while it is running at our DR site. I can build a new NDMP node, but it will take about a week for the backups to finish since the new node would not have any of the cached files to reference and would have to scan the whole 8TB of data again.
I am trying to re-register my file server with the current prod NDMP node, in theory there shouldn't be a whole lot of data traversing our link between prod and dr, but I would still like to have an NDMP box at the DR site in case we were not able to use the prod NDMP node.
or if anyone has had this before and had other ideas, I am open to those as well.
Solved! Go to Solution.
NDMP doesn't use the file cache so the benefit you'd get from the caches would be relatively small. Instead of using the file cache, Avamar NDMP relies on having a previous backup available to use as a base and asking the array for the list of files that have changed since that backup was created.
The NDMP client retrieves a list of recent backups from Avamar and records the dump date of the most recent matching backup. When the NDMP client sends the L1 backup request to the VNX, it uses this dump date to request only the files that have changed since the previous backup.
In 6.1 and older versions of Avamar, there is an issue with backing up arrays that are replication targets due to the way replication interacts with modification dates -- it's possible for replication to break a file's history chain in a way that causes backups to fail. This is resolved in 7.0 as long as there are no files on the array where the modification date is in the future.
we are running version 7, so it is possible that I can just setup a new ndmp node at our DR site, and just run a backup and it won't take that long? If I am reading what you wrote write, or hoping that would be the case in my mind
Unfortunately I don't think that will work.
I spoke with one of my colleagues on this. Even if we copied the backup into the client account for the accelerator at the DR site and if the client could be convinced to use that backup as the base for an incremental (which is a big "if"), the inode numbers on the source and target would have to be exactly the same for every file for the backup to succeed. I don't think VNX replication preserves inode numbers.
I think the best you'll be able to do is copy the p_cache.dat file from the source accelerator to the DR accelerator and run a level 0 of the DR target array. Most of the data will already be on the Avamar server so it will be a relatively fast L0 but it will still be an L0.
I will give that a shot and see what we get. As long as I can get backups a some point in time, if I were able to just copy some info over and just start off from the last backup that would have been great. But I guess sometimes you can't have your cake and eat it to!
thanks for your help.
Once you get over the L0, the cost of keeping backups of the DR target running is relatively low -- they won't consume additional capacity on the Avamar server. Probably best to just keep them running in case you find yourself in this situation again.