Protect stgpool that is also used for replication destination

[smp]

Here are the simplified steps that I want to perform.

TSM client NodeA backs up data to ServerA StgpoolA.
TSM client NodeB backs up data to ServerB StgpoolB.

Step1(Without replication configured):

ServerA:
StgpoolA
NodeA

ServerB:
StgpoolB
NodeB

Step2(With replication from source(ServerA) to target(ServerB))
ServerA:
Protect stgpoolA
replicate nodeA
NodeA(syncsend)

Results:
ServerB:
All extents protected to stgpoolB
nodeA meta data updated on stgpoolB
NodeA(syncreceive)
NodeB

Step3:
Configure replication from source(ServerB) to target(ServerA)
ServerB:
Protect stgpoolB
replicate nodeB
NodeB(syncsend)
NodeA(syncreceive)

Results:
ServerA:
All extents protected to stgpoolA
nodeB meta data updated on stgpoolA
NodeA(syncsend)
NodeB(syncreceive)

Questions:
1. Is it ok to have a single storage pool used for both TSM client backup data and replicated data from another server?
2. During step3 protect stgpoolB, will the system know that the only extents that need to be updated on stgpoolA are extents
that were added as a result of the backup of NodeB on ServerB?
3. Would there be any issues with the protect stgpool running on both servers simultaneously?

 

[marclant]

1. Is it ok to have a single storage pool used for both TSM client backup data and replicated data from another server?

Yes.

2. During step3 protect stgpoolB, will the system know that the only extents that need to be updated on stgpoolA are extents
that were added as a result of the backup of NodeB on ServerB?

It is my understanding that you can. Protect stgpool will copy from one server to the other the extents that don't already exist on the other side. It doesn't matter who owns the extent, you can have an extent that is both owned by nodes from both server because they are both backing up the same files. Think of OS files, they are common.

3. Would there be any issues with the protect stgpool running on both servers simultaneous

Yes, but you have to be very conscious of the number of sessions. I don't know what Blueprint size you fall under, let's assume a Medium. A medium can handle 40 concurrent sessions in total. So if you run both simultaneously, you will want to use 20 sessions each. If they are one after the other, than you can do 40 sessions each. Same applies for replication or protect in one direction and replication in the other. You want to keep the number of concurrent sessions of protect and replication to 40 at any given time in all directions.

Also, when you setup for cross-replication, your servers have to be sized double what they would normally be. For example, if you have a server that has a 25 TB daily ingest (client backup), and the other 20 TB. If you cross replicate, they are both receiving 45 TB as the replicated data received is processed like client backups received, so that means you need 2 large servers instead of 2 medium servers.

 

[smp]

Thank you so much for the detailed response, marclant. This is very helpful.

 

[smp]

Yes.

It is my understanding that you can. Protect stgpool will copy from one server to the other the extents that don't already exist on the other side. It doesn't matter who owns the extent, you can have an extent that is both owned by nodes from both server because they are both backing up the same files. Think of OS files, they are common.

Yes, but you have to be very conscious of the number of sessions. I don't know what Blueprint size you fall under, let's assume a Medium. A medium can handle 40 concurrent sessions in total. So if you run both simultaneously, you will want to use 20 sessions each. If they are one after the other, than you can do 40 sessions each. Same applies for replication or protect in one direction and replication in the other. You want to keep the number of concurrent sessions of protect and replication to 40 at any given time in all directions.

Also, when you setup for cross-replication, your servers have to be sized double what they would normally be. For example, if you have a server that has a 25 TB daily ingest (client backup), and the other 20 TB. If you cross replicate, they are both receiving 45 TB as the replicated data received is processed like client backups received, so that means you need 2 large servers instead of 2 medium servers.
View attachment 1764

Just so I can better understand, suppose I am only running replication one-way to a storage poolB that only contains replicated data. Example, target serverA has 10 TB of data in storage poolA that is replicated to an empty storage poolB on the target serverB. Both storage pools now contain 10 TB of data.
Questions:
1. If I don't add anything else to storage poolB and then run protect stgpoolb and replicate node * back to serverA, what is the total amount of data now on storage poolA?
2. Will the total storage used on poolA and poolB remain the same since they were exact copies of each other before I ran the replication back to ServerA from ServerB?

 

[marclant]

In the scenario you just posted, replication on ServerB won't run because the nodes are setup to receive on serverB, not send. If you have nodes on ServerB setup to replicate to A, than only those nodes would replicate, but that would add to ServerA.

But, what you really want to know is if you use cross-replicate as in your original post where you have nodes on both A and B doing backups, and replicating to the other server. If you only have 1 pool on both servers, doesn't matter if you go from A to B, or B to A, when protect or replication runs, it will not copy an extent to the other server if the extent already exists in the storage pool of the other server. In the case of protect, it will move on to the next extents, because protect only deals with extents. In the case of replication, same thing, it will not copy the extent from serverA to serverB (or vice-versa), if the extent already exists on the other side, instead it will add a reference that this extent belongs to the node/file being replicated.

A unique extent is only stored once and referenced multiple times if it belongs to multiple files regardless if this is on the source or target server.