CitectSCADA and VMWare snapshot backup

Hello everyone,

We are using CitectSCADA 2016 with two redundant servers. Each server is running in a virtual machine (VMWare). Everything works fine except that sometimes (about once or twice a week), the backup server takes control for a few seconds (the IODevices switch to the standby) and then the go back to the primary server. Since this always happens between 22h30 and 23h00, we are trying to find what causes this.

After discussing with our IT department which manage the servers, they do a backup of the VM machines at about those times everyday by doing a "snapshot" of the virtual machine. We are wondering if this could cause a "glitch" on the primary server when this backup is made, which would cause the standby server to not see it for a short period of time and take control.

Has anyone had any similar experiences with VMWare, are there any "best practices" concerning VM configurations and backups when using Citect ?

Thanks for your help,

Patrice Jacob

Prosystech inc.

Parents
  • I believe the underlying issue is that VMWare puts the VM in a halted state during the backup (possibly using the Windows Volume Shadow Copy Service in the guest-OS). If taking the snapshot takes too long, the redundant server will lose connection to the primary and will become active until the backup has finished.

    I'm not a VMWare expert but Google finds that the "Quiesce" option might be related to this. Toggling this option might have some unwanted side-effects however, especially on domain controllers and machines that run databases, if I understand the article correctly.
Reply
  • I believe the underlying issue is that VMWare puts the VM in a halted state during the backup (possibly using the Windows Volume Shadow Copy Service in the guest-OS). If taking the snapshot takes too long, the redundant server will lose connection to the primary and will become active until the backup has finished.

    I'm not a VMWare expert but Google finds that the "Quiesce" option might be related to this. Toggling this option might have some unwanted side-effects however, especially on domain controllers and machines that run databases, if I understand the article correctly.
Children
No Data