CRAZY slow replication between 2x 2016 hosts

[mvalpreda]

Veeam 9.5u3 on 2x Windows Server 2016 hosts

Replication jobs are CRAZY slow. Single 900GB VM with ~560GB of data took nearly 8.5 hours. Having the same issue with another couple of VMs and the job has been running for just over 11 hours and has only processed 288GB and transferred 167GB. Veeam reports the bottleneck as the network. Throughput graphs show quick bursts of data transfer with then 30-60 seconds of no data moving.

Machines are on the same switch, no VLANs. If I copy a large file from one host to the other and vice versa I get full gigabit wire speed. Backups from either machine to a Synology over iSCSI get 100mb/sec. Only replication jobs are slow slow slow slow slow.

Case 2453689

[Mike Resseler]

Hi Mark,

thanks for posting the case number. Please keep working with our engineers. Any change you have throttling rules between the two hosts and/or data mover components of Veeam?

Cheers
Mike

[mvalpreda]

No rules at all. Just a vanilla install of Veeam and then added the replication target. Done this a lot of times before, just never seen it run so slow.

[Mike Resseler]

Mark,

Since you already have a case number, please keep working with the engineers. Hopefully they can see in the logs what is going on

Mike

[mvalpreda]

Interesting.....I saw SQL was using ~3.5GB of RAM. Cut that down to 1GB through SQL Management Studio and now it is working much faster.

[Mike Resseler]

As in our SQL? Or a SQL server that you are replicating?

[mvalpreda]

Sorry.....yes....SQL Express for Veeam. Connected to the database instance and set the memory to 1024 from unlimited and it is running a lot better. Seemed like a strange fix.

[Mike Resseler]

I notice that your case is still running so please continue with it. I believe the engineer has requested to run a few tools for measurements.

For that SQL thing. It is not really strange. SQL itself is known to "eat" all the memory it can get. (For example, if you have a VM that can grow its memory at request and that is running a SQL box... ). But it seems that in your case when it takes that memory (although only 3.5 GB) the system starts to slow-down. Does the B&R server has enough RAM? But it could also be an underlying disk issue.

Anyway, I hope that our engineers find something after the results of the tools
Mike