Slow replica seeding

[fre1621]

I been struggling with seeding a remote site to our Data Center.
In version 5 the process was very simple: create your replication at the remote site to local storage (UBS drive) then physically take to DC, copy the data to the correct datastore and then kick off the replication job again. This worked ok, and we saw throughput of around 10MB/s over our VPN's T1 connection.

With version 6 I have been seeing very slow initial seeding throughput of around 140Kb/s which is around the speed of the T1.
I have configured it per another post:

1. Configure a backup job for the VM you want to perform initial seeding for.
2. Create and choose any local repository for the backup files.
3. Once you have backup files created, move them to a DR site and place them on the DR site repository (you should create it beforehand).
4. Scan DR site repository to populate backup files in the backup console.
5. After that just create a replication job with "Low connection bandwidth" checkbox enabled.
6. On the "Seeding" step of the replication job wizard specify DR site repository where you've placed VM backup files.
7. Finish configuration of replication job wizard and start replication job.

I know that the initial seeding will take longer but once it's done the actual replication job will speed up.

I have talked to tech-support. They confirmed that I had the job setup correctly and that this is normal for seeding under version 6 and could take days.

At this rate it will take me about 9 days to get my initial seed, which is about 150GB of data. Just not acceptable

Has anyone else seen these kind of results?

--
Thanks, Fred

[Gostev]

Do you have a backup proxy server in your DR site, and do you have it selected as the target proxy in the replication job settings? In that case, only compressed delta between actual and backed up state will be transferred over WAN as a part of replica seeding.

By the way, v5 did not have a "throughput" counter, this is a new v6 feature. Of course, you cannot have 10MB/s throughput over T1 link, so most likely you are comparing apple to oranges here.

[fre1621]

Yes I do have a proxy (win XP, two processrors, 4GB ram). And it is set as the target.

While yes I know that I can't have a T1 @ that speed, that was the pould rocessing speed. However, why would it go at that speed if I physically copied it to our DC anyway?

[fre1621]

Yes, I am storing the metadata at the source (remote) site and the seed data is in a repository at the DR site. I have also noticed that I am getting failures about 12 to 15hrs in to the job. Error "Veeam.Backup.AgentProvider.AgentClosedException' was thrown". Sent the logs to Tech support, but no insight yet.

[Gostev]

While yes I know that I can't have a T1 @ that speed, that was the pould rocessing speed. However, why would it go at that speed if I physically copied it to our DC anyway?

This is exactly why I asked about the second proxy at the target site. It does not look like it is being used at all, so may be you don't have something configured correctly. Naturally, it would be best to let our support engineers take a look at your configuration over the webex.

[fre1621]

They assured me that this is normal speed for the initial seeding (seemed odd to me). We did review the settings and it seemed to be correct. I just sent them my latest logs that might shed some additional light on this. Will keep you posted.

[Gostev]

What is your support case number?

[fre1621]

What is your support case number?

ID#5177301

--
thanks, fred

[creativeTechie]

I have a similar problem. Did you get this resolved? Thanks. Matt

[foggy]

Matt, there are two issues discussed in this thread actually - the speed issue and an exception thrown issue. Which one do you experience?

[fre1621]

I have a similar problem. Did you get this resolved? Thanks. Matt

I found that the trick was my Repository that was in my Data Center needed to be processed using my Data Center Proxy. The default on the Repository Share was set to "directly from backup proxy" Change it to point directly to my Data Center Proxy and the speed was much, much better.

The initial seeding still takes a long time, but once it's done my normal nightly Replication job is at least twice as quick than it was on version 5.

The other issue we had was that the job would quit many hours in to it. We found that the RPC service was off. Turned on again and with the change I mentioned above I was able to get the job finished.

--
Fred

[skeeziix]

This does indeed work. Make a proxy of the source and destination... specify that in the replication job from source to destination. Thanks!