Vm freeze during replica

[Jneau]

When a replica is in progress, Vms freeze for at least 4 min, the duration of the replica, and all users are disconnected. No error in the job replica, the vmtools to date. I tested with Vss enabled or disabled, the same thing. No latency found in the logs ESX and vCenter

config:
ESXi 5.1 U2
Veeam 7.0 build 839
Vm 2008 standard
Case : 00559591

Thank you for your reply,

[foggy]

Could you please elaborate, does the VM experience the freeze during the entire job duration or just at the snapshot commit stage? What kind of VM it is (does it run any highly transactional app)?

[Jneau]

The vms freezes for the replica, to the creation and deletion of snapshot. The Vms in question have 64-bit Sql 2008 and a standard application rds / tse windows server 2008.

I tried activating and deactivating Vss on vms via veeam job.

Thx,

[veremin]

I'm wondering whether this issue can be reproduced without Veeam presence. You can take the snapshot of VM manually, keep the snapshot open for long enough (similar to time it takes to replicate the VM), delete the snapshot and see whether the freeze problem shows up.

Thanks.

[Jneau]

Below the response of my client: (wait 15 min between start and delete sanpshot)

When creating the snapshot, I found no freeze.

When you delete the snapshot, there has been a slight freeze two seconds on the VM TSE02. However, the snapshot is deleted extremely quickly compared to do with Veeam.

[Vitaliy S.]

What did our engineer say on this? Is it only replication job that is affected or backup jobs also experience the same behavior? How long does it take to run a replication job against this VM?

[ksauer]

VM snapshot removal stun is expected behavior of any Virtual Machine in VMware since VMware has to shift the active writes to disk from the snapshot file back to the base disks. This is more common and quite troublesome in situations where heavy I/O servers (like Exchange for example) live on NFS datastores. This is explained in detail in the following KB article:

VM loses connection during snapshot removal
http://www.veeam.com/kb_articles.html/kb1681

As indicated in the article there are several changes that can be made in the infrastructure to reduce the network downtime of a VM during snapshot removal. VMFS datastores tend to handle this operation much more smoothly than NFS because VMware developed the file system and has more active control over it where with NFS they are borrowing an existing file system so the file locks/unlocks aren't as fast they could be which sometimes results in these periods of downtime. I have seen this on VMFS datastores as well, but the downtime is significantly less. This behavior is highly reproduce-able without Veeam's involvement as you can typically take a manual snapshot in VMware, let it sit for the same duration as it would sit from a backup job (so it can grow in size), and then manually remove it. You would notice the same behavior in this test with your servers losing network connection.

Although this is certainly more common with NFS datastores, it can certainly happen with VMFS if the VM is under heavy load.

Hope this helps!