"Automatic" failover to Non-vCenter-Backup

[timmi2704]

Hi guys,

I have a question regarding the backup process without a vCenter Server.
I know about the disadvantages of backing up from the ESX-Hosts directly (vMotion,...).
But I need to work out a possibility to automatically Failover from a vCenter-supported-Backup to a backup directly from the ESX-Hosts in case the vCenter is unavailable.

I thought of something like a powershell script which runs as a scheduled task at the beginning of the backup time window.
1. Check whether vCenter is available
1a. If Yes, end script. Usual scheduled backup Job in Veeam will run
1b. If no, continue with script
2. Start a preconfigured disabled backup-Job, which does not use the vCenter, but uses the Hosts directly

What are your thoughts regarding this idea?
Is there a better way to realise my needs?

Any answers regarding this are greatly appreciated
Thank you.

[veremin]

With Powershell it should be pretty achievable task.

However, the challenge would be false alarms. For instance, vCenter unavailability might be caused by variety of reasons, even the network hiccup. In this case, the script would be executed, see that vCenter is currently down (even if it's not true) and execute the Host-specific replication job, which, in its turn, will result in undesired consequences.

Thanks.

[timmi2704]

That's a good point, thanks.

I think I'll try using the VMware vSphere Powershell Plugin to avoid false positives.

If I set up a Job where I add the ESX-Hosts only, will Veeam ask the Hosts which VMs remain on it every time the Job is started?
I think this is the usual behaviour, isn't it?

Also I read, that the VMs will have a different ID. So the Job will perform a full backup even if I select the same target repository?

[veremin]

If I set up a Job where I add the ESX-Hosts only, will Veeam ask the Hosts which VMs remain on it every time the Job is started?

You mean select a given host as source for a backup/replication job, right? If so, then, yes, only the VMs that currently reside on the host will be backed up/replicated.

Also I read, that the VMs will have a different ID

What VMs you're talking about? VMs added via vCenter and via standalone host?

Thanks.

[timmi2704]

You mean select a given host as source for a backup/replication job, right? If so, then, yes, only the VMs that currently reside on the host will be backed up/replicated.

That's exactly what I meant - Thanks.

What VMs you're talking about? VMs added via vCenter and via standalone host?

You are right. Assuming I have two Backup Jobs, and they both back up the same VMs to the same backup repository - One Job over the vCenter, and one Job directly over the Hosts - will the selected VMs have different IDs, or will Veeam recognize that These are the same VMs and therefore do an incremental backup?

[veremin]

The backup jobs are completely independent on each other, therefore, they will act separately, according to their own settings. Thanks.

[Vitaliy S.]

I think I'll try using the VMware vSphere Powershell Plugin to avoid false positives.

Just out of curiosity - how are you trying to avoid false positive via VMware vSphere Powershell Plugin?

[timmi2704]

I am not sure yet.
As I have no advanced VMware knowledge it might be a combination of some basic tests.
E.g. first a network availability test, then a service availability test.
After that I'll try to get the vCenter object with the VMware powershell snapin.

This would be my first idea, but if anyone has some additional or even better ideas, I would be very greatful.

[Vitaliy S.]

You definitely need to verify multiple sources before failing over to host direct connection. In addition to what you've listed above, I would recommend including some kind of a timeout in order to be protected from network glitches.

[veremin]

Anyway, however many parameters you use, you won't be solely protected against accidental false alarms. So, in the end it comes to weighting pros/cons of this solution.

Thanks.

[timmi2704]

I think both of you are right. Even if I try to use several sources I will never absolutely sure.

A live test will show, if the following works:
1) Checking network availability, if unavailable wait 10 minutes, then try again.
2) Checking status of vCenter Service
3) Try to get vCenter Object with VMware snapin. Check, if gathered object is valid.
If any of the above mentioned steps fails (Second network check fails, service status is other then "running", gathered object is invalid), a prepared unscheduled job will be triggered to start to backup the hosts directly.

Thanks for your impulses.