Veeam Replication vs Veeam Backup snapshot technology

[ckbrou]

We are looking at some different DR options (Zerto, array based replication, and Veeam Replication). We love Veeam for our backup needs and are seriously considering it for this DR / Replication project. I am having a hard time getting information from our vendor, though, on how Veeam replication works from a snapshotting standpoint. Does Veeam replication also utilize vmware snapshots just like Veeam backup? We have a NetApp array with snapshot integration (which works great!) but we still lose a ping or two when the vmware snapshot takes place and when it is removed. This is fine for backup jobs since they take place outside of business hours; but that would not be the case if we were to rely on Veeam for hourly replication. Our ERP system is very sensitive to IO freeze on the DB server and dropping even a ping or two can cause issues.

I believe with Veeam backup I can test IO freeze / ping loss by taking a snapshot from vcenter (uncheck "snapshot the virtual machine's memory" and check "quiesce guest file system"). Is this a good test for Veeam replication also?

[dellock6]

Hi Chad, it is correct both that we use vSphere snapshot technology to do replicas as we do for backups, and also that you can leverage NetApp storage snapshots to initiate a replica at source to lower their impact. Have you done any research on the performance to figure out why the loss of connection is happening even when using storage snapshots? There may be some performance issue that can be improved.

For the test, it's almost correct, but we do not snapshot memory, so you do not need to select any of those two options. A suggestion can be to look at the backup job usins already storage snapshot, or even look at vcenter during a backup, and see the time it takes between snashot creation and commit. This is the time it will take even during a replication. Once you know that value (probably around one minute) you can test the impact of a replica during the day by going into vcenter, take a snapshot of the ERP machine, wait for the same amount of time, and then ask vcenter to delete the snapshot. Look if the ERP machine is having any freeze or slowdown.

Luca

[ckbrou]

Thanks for the quick and useful response! So is Veeam Replication just crash-consistent then since you mentioned (for the test) unchecking both "snapshot the virtual machine's memory" and "quiesce guest file system"?

[dellock6]

No, it's also application consistent, but with vcenter snapshots you cannot simulate the VSS interaction or any scripting activity like in a real Veeam backup. We use a different technology, not the one offered by vmware tools.

[ckbrou]

Okay great! I can do some more research to see where the performance bottleneck is. I thought it was "expected" though to always lose 1 or 2 pings when taking a snapshot and when removing the snapshot. Is that not the case anymore?

I am on vsphere 6.0, new UCS blades for hosts, and a fairly new NetApp FAS array using iSCSI protocol.

[foggy]

Lost pings during snapshot commit are typically expected depending on the storage performance and how active VM is. However, things have changed with vSphere 6.0.

[ckbrou]

Thanks for linking to that post... so it looks like dropping 1 or 2 pings with vSphere 6 is normal snapshot commit (much better than vSphere 5.5 in our experience also). We will just have to decide if that is acceptable for our ERP application during business hours.

Love this forum. I always get quick and practical responses!

[meetoo]

We are testing Veeam 9.5U2 & during an initial backup of a Vmware 6.0 VM, we had 2 ping drops.

Is this normal behavior?
Is the number of drops related to the size of the VM? We have several VMs that are more than 3TB.

This may be a deal breaker.....

Thx

[PTide]

Hi,

That behaviour is normal in some situations, please see foggy's response and review "Snapshot removal" section of this document for more info. Please refer to this KB for troubleshooting tasks.

Thank you