Slow Replica creation

[hampi]

Hello
I installed an ESXi (VMWARE Essentials License) yesterday. The Server has 2 Datastores configured (each composed out fo 2 SATA Disks with Raid1). vCenter and Veeam (2 Socket Standard License) are installed together on a Windows Server 2008R2 64bit VM.
For testing, i created a replica job for a running Windows XP VM with 13GB of Disk. The replica was created on the same ESXi but on the second Datastore.
In network mode it took nearly 5 hours to create the replica. In the Vitual Appliance Mode it took even longer.

The Hardware is a brand new Dell T310 with 4 Core Xeon and 8GB of Memory.

Is this duration normal? Any suggestions to improve?

Regards
Hans-Peter

[Vitaliy S.]

Hans-Peter,

Could you please use vSphere Client -> Datastore Browser and upload any big file from the backup VM to target datastore? What are your speed rates?

Thanks.

[hampi]

Hello Vitaliy

I tried as you told and i was able to copy 367MB in about 46 seconds. Here are the data rates i found in the logs:

• Copied 367MB via Datastore Browser: 7,7MBytes/s
  Replica of 40GB VM (Appliance Mode): 2MBytes/s
  Replica of 13GB VM (Network Mode): 813KBytes/s

Regards
Hans-Peter

[Vitaliy S.]

The processing rate you see in Veeam Backup console is not the actual transfer speed. The processing rate is calculated using this formula: Total VM size/Time (and so is affected by time required to perform guest processing, perform guest freeze, create and remove snapshots, backup smaller files). So the throughtput looks comparable based on these numbers. That being said, I suggest looking through the corresponding backup job log (Help|Support Information) to get the real numbers of transfer rate.

It is expected to have not high speed while using ESXi as target for replication, because ESXi host uses agentless mode to transmit all the data, see this topic for more information: Slow Replica

[Vitaliy S.]

Hans-Peter, have you tried running an incremental job pass? What figures do you get in this case?

[hampi]

Hello
I tried the incremental job and i found a processing rate of 18MBps. But the running time was 38 minutes for a job on a server with no real changes. Quite a lot.
Regards
Hans-Peter

[Vitaliy S.]

Hans-Peter,

Reading back through my response above, please check out the corresponding job log (you may contact support for assistance) for the number of changes that have been processed.

Veeam Backup VRB/VIB size is equal to the amount of VMDK blocks changed since last backup, so generally you need to look for reasons what is causing so many disk block changes. Please keep in mind that defragmentation/antivirus scanning might also cause block changes.

Thank you!

[hampi]

Hello
I managed to get better performace by optimizing my infrastructure. I added a iscsi target (1Gbit Network) to the Backupserver and was able to get a processing rate of 37MB/s for the initial backup. Incrementals are much faster. Thats ok for me.

I still have one problem:
When i do a replication of a VM (backup processed with 37MB/s to iscsi target) to a datastore on a second vm-host, the performance is still very poor.
Example: VM is located on Datastore1 of VMHOST1. The Backupserver with Veeam B&R installed runs on the same VMHOST1. The replication target is on Datastore1/VMHOST2. The hosts are connected by a 1GBit Network.
Result: The statistik shows that 9.71KB are processed and then everything stops. I waited for more than 30 Minutes but nothing happend. Remaining Time was showed as about > 18 hours. I tried to cancel the job but was not able to do so. After a server restart, the job had stopped.
Questions:
- Can this be normal?
- What type of replication (virtual appliance, network, usw) would you suggest for this environment?

Thanks again
Hans-Peter

[Vitaliy S.]

No, that should be normal. If the job processing rate drops to zero or even stops, I suggest reviewing the corresponding job log, as it seems like you might have some connectivity/timeout issues. You may contact our support team for assistance here.

By the way what is the replicated VM in question? Is it an Exchange Server or are we still talking about WinXP VM with 13 GB hard drive?

Network mode is the slowest mode comparing to other vStorage API modes, so I would recommend choosing Virtual Appliance mode (please make sure you have 4 vCPU on the backup server VM).

[Vitaliy S.]

If you don't have any problems with backing up this VM, then the bottleneck should be in destination storage (ESXi host), moreover you should see the same behaviour for other VMs... Is that true? I think it's the right time to check out the replication job log file to shed some light on this behaviour.

[Gostev]

Hi Hans, I have deleted your post containing huge log file to keep this topic readable. Please review the forum rules regarding posting about technical issues provided when you click New Topic. Thanks!

[hampi]

Ok. I sent a mail to the support.
Hope it helps.
Regards
Hans-Peter