I have a customer who is using Veeam to migrate some VMs from one Vcenter environment to another. He has set up two virtual proxies, one on the source Vcenter and a second on the Destination Vcenter. Each Proxy VM has 24GB of RAM and 16 vCPU’s. He has set Max concurrent tasks to 20 on both proxies and has a green check mark on that. He set up a job with multiple VM’s with each VM having 2 disks. The job is only processing a max of 4 VM’s at at time. I am thinking it should be able to do at least 10 VM’s (10 x 2 disks = 20 concurrent tasks). Do VM replication jobs use more concurrent tasks than a single task per disk?
Page 1 / 1
Yes, technically you should be processing more tasks than what you are seeing based on the task count per proxy server.
See the note on this page - Limitation of Concurrent Tasks - User Guide for VMware vSphere
How many hosts are in each cluster? They may be better deploying smaller proxies with one on each host.
Hmm….interesting. There could be a couple things at play here, but may be best to confirm with Veeam Support. Replication may do concurrent tasks as a “per Job” option like Backups which don’t use per-VM Repos. There could be one thing going on. Also, and I’m not entirely sure on this theory...but Veeam could also be detecting latency on the Datastores and doing throttling. But not sure if that reduces tasks alongside the Concurrent Task Limit setting.
I’d recommend a technical question like this to also be asked in the R&D forums as many Veeam employees frequent it and give great advice.
Here is some info on Replication.
https://helpcenter.veeam.com/docs/backup/vsphere/replication_process.html?ver=120
Veeam Backup & Replication performs VM replication in the following way:
- When a new replication job session starts, Veeam Backup & Replication reads job settings from the configuration database and creates a list of VMs to process. For every disk of a VM added to the job, Veeam Backup & Replication creates a new task.
So using Veeam replication I’d agree 10 VM’s with 2 disks each should be correct.
If he is using the Repository to seed the backup, there is a task setting on the Repository as well.
https://helpcenter.veeam.com/docs/backup/vsphere/replica_create_seed.html?ver=120
If you are using a VM, id try and put one on each host, or cluster at a minimum.
And if you are seeding, 1 VM = 1 task if using Per-VM chains, and 1 JOB = 1 Task if not using Per-VM chains.
This is probably silly, but maybe the problem is in the task limit of the repository used for replica metadata? 
I did post on the R&D Forums and I believe I have the answer. The default number of concurrent VMware snapshots that Veeam is allowed to take per datastore during backup/replication is 4. This gate is in place to ensure that Veeam does not overrun their VSphere environment with VM snapshots. There is a registry setting that can be added to up this value. Once the customer added the registry setting and changed it to 6 then he saw 6 concurrent replication sessions running instead of 4. Most of the current VM’s he is trying to replicate are on the same datastore. The recommendation is to increase the value in small steps to avoid performance issues on their VMware environment.
Here is a link to the details of this setting:
https://bp.veeam.com/vbr/3_Build_struct ... -snapshots
Great to hear you received an answer for your query. Thanks for sharing with the community.
Ah, ok...thanks for sharing
Comment
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.