Vault Duplication of Images on Dedupe Pool

[kenrblan]

My environment is Netbackup 7.0.1 on Windows 2008 R2. I have a master server plus three media servers. Two of those media servers have Deduplication Disk Pools on them. One of those is in our production site and is where backups are initially done. I have a vault job that performs an optimized duplication of specific servers' backups to the other media server located in our DR facility.That part of the setup is working.

My problem is that I also need to duplicate images to LTO-4 tape since we don't have the bandwidth required to replicate the entire Dedupe pool to DR. It appears that the Vault duplication to tape is trying to read production site deduped images with the DR site media server. That is, of course, slowing down my tape duplication jobs. It doesn't make sense that Netbackup would try to read the image from the other media server when the source image resides on the same server that is writing the tape.

I am including information from the details tab on one of the Duplication jobs:

3/1/2011 3:22:22 PM - begin Duplicate
3/1/2011 3:22:23 PM - requesting resource bxsops078-nb002-hcart-robot-tld-1
3/1/2011 3:22:23 PM - requesting resource @aaaa3
3/1/2011 3:22:23 PM - reserving resource @aaaa3
3/1/2011 3:22:24 PM - reserved resource @aaaa3
3/1/2011 3:22:24 PM - granted resource L40011
3/1/2011 3:22:24 PM - granted resource IBM.ULTRIUM-TD4.hcart.003
3/1/2011 3:22:24 PM - granted resource bxsops078-nb002-hcart-robot-tld-1
3/1/2011 3:22:24 PM - granted resource MediaID=@aaaa3;DiskVolume=PureDiskVolume;DiskPool=BXS-DedupePool-1;Path=PureDiskVolume;StorageServer=bxsops078-nb002;MediaServer=bxsdrc078-nb002
3/1/2011 3:22:27 PM - started process bptm (5956)
3/1/2011 3:22:33 PM - started process bpdm (6580)
3/1/2011 3:22:34 PM - started process bptm (5956)
3/1/2011 3:22:34 PM - mounting L40011
3/1/2011 3:22:34 PM - started process bptm (7952)
3/1/2011 3:22:36 PM - started process bpdm (2724)
3/1/2011 3:22:46 PM - begin reading

In the Storage Unit settings for the Production Site Deduped Disk, the "only use the following media servers" option is selected and set to only use the production site media server.

Any ideas on what is going on here?

[Marianne · Accepted solution]

I don't think Vault can be configured to use duplicate images as source - it uses the original backup copy as source.

Consider using SLP's instead where you can determine which image must be used for duplication.

See example below:

[kenrblan]

In this instance, the primary image (original backup copy) resides on the production (OPS) media server which is also the media server that hosts the tape library that is performing the Vault duplication. I don't have any original backup copies on the DR media server, but that media server is setup to perform optimized duplication of some images thus has access to the dedupe pool.

When my tape vault job runs, it is trying to use the DR media server to read the images stored on the production (OPS) storage that are being written to tape on the production (OPS)  media server. Since that is over a slow link, I am getting an insane bottleneck where one should not exist. The production (OPS) media server should be reading the images from the dedupe pool and streaming it to the local LTO-4 library. The traffic should never leave the OPS server.

I suspect that I may need to remove the DR media server as a load-balancer for the production dedupe pool, but keep the production (OPS) media server as a participant in the DR pool so that I can still replicate to DR.

[kenrblan]

I unchecked the DR media server from the Credentials->Storage Servers Puredisk media servers properties for the primary site storage server. That fixed the slow duplication to tape problem, but it broke optimized duplication from production to DR. I started getting  "NBU status: 2101, EMM status: Media server not found in EMM database" errors when optimized duplication ran. Re-selecting the DR media server for the production storage settings fixed the error and optimized duplication is able to proceed. Any ideas on how to get both things working properly?