Replication

[ashman70]

I currently have two hosts that I plan to upgrade very soon, both are running Esxi 5.1u3. At the moment, the replica host has a predictive hard drive failure so I want to replace the host with a newer server that have. I don't want to put 5.1U3 on it, so can I put 5.5U3 and still be able to replicate without any issues? To be clear the replication goes from the server running 5.1U3 to the new one running 5.5U3

[Shestakov]

Hello,
Replication between Esxi 5.1u3 and 5.5u3 should work with no issues.
Thanks!

[ashman70]

Thanks. I setup the new server which actually has a 10 core processor whereas the primary vmware server has an 8 core processor.

For some strange reason, the replication seems slower with the new server than the old. Any particular reason for this? I am using the same network drops as the old replica host.

An old replica job has these statistics:

Hard disk 2 (320.0GB) 258.0 MB read at 26MB/s [CBT]
Hard disk 1 (80.0GB) 116.0 MB read at 20 MB/s [CBT]
Busy:Source 99% > Proxy 13% > Network 8% > Target 27%
Primary bottleneck: Source
Processing rate: 37MB/s

A new replica job for the same VM has these statistics:

Hard disk 2 (320.0GB) 216.0 MB read at 14 MB/s [CBT]
Hard disk 1 (80.0GB) 117.0 MB read at 10 MB/s [CBT]
Busy:Source 49% > Proxy 23% > Network 5% > Target 99%
Primary bottleneck: Target
Processing rate: 18MB/s

Why the discrepancies?

The old host had older hardware and was running Esxi 5.1 the new host has more powerful hardware and is running Esxi 5.5U3

Is this a Vmware configuration issue?

[Shestakov]

According to the statistics, your bottleneck was Source, now it became "Target", i.e. target (replication storage) disk writer component.
Your target storage speed is presenting a bottleneck for the whole data processing conveyor, because all the pending I/O operations cannot complete fast enough, and due to that there is always some data waiting in the incoming queue of the network component that is waiting to be written to disk.
Have you changed a datastore after the upgrade?

[ashman70]

Yes the replica server was replaced, storage is local. On the old replica server there were six 900GB SAS drives in RAID 10, on the new server there are four 1.8TB SAS drives in a RAID 10. The old replica server had a Dell Perc 6i, the new replica server has a higher end lenovo RAID controller with 1GB NV cache memory.

[Shestakov]

Check if no other IO operations are made during the job activity. You can leverage Veeam ONE Monitor for that purpose.
I would also check if all replication job runs indicates Target as a bottleneck.
Thanks!