VM network connectivities lost when taking a Snapshot for backup

[hpoirier8]

Hi,

On a production environment, when the backup tool triggers its snapshots creation, that carries out to the lost of some Network connection on a specific VM. The different Network interfaces are configured with VMXNET3 adapters;

VMware sphere version on host is "VMware ESXi 6.0.0 build-8934903"

is there a way to confirm that there is a link between taking a snapshot and the network connection lost, except that it is at exactly the same time ?

Is hthis a known issue already encountered on other environments?

Thanks for your update.

[e_espinel]

Hello.
Version 6.0 is very old and is out of support in VMware. You should consider upgrading to 6.5 or 6.7 if your hardware supports it.

VMware Compatibility Guide- HW

http://partnerweb.vmware.com/comp_guide2/search.php

Are the VM's you want to backup on the server's internal storage or on some external storage?

You are using the physical backup method, i.e. using agents and using the network card. Is this your method?

In general it is recommended to separate the network traffic but it depends on the amount of ethernet cards we have in the server.
One ethernet card for the ESXi administration, another one for the Data (VM) and another one for the backup would be ideal.

 

 

[hpoirier8]

Hi,

first of all, thanks for your answer.

Regarding Upgrade, there is an upgrade planned in the coming months.

The design that has been favored for this information system is not necessarily optimal, but it is supposed to work. We have 2 subnets configured, one for OaM (Operation & Management) and a second one dedicated to applications needs. Backups rely on OaM Subnet.

But the question is : Do you think that there could be a causal link between taking a snapshot for the backups (Rubrik backup solution), and the loss of the network connections? is it something that has been already seen ?

Thanks in advance for your update

Rgds

hubert

[e_espinel]

Hello.
If I understand you correctly, you have one ethernet card for Admin and backup, and another ethernet card for Applications (Data VMs). then you should also have vlan to separate the admin and backup traffic.

The speed of the ethernet card should match the speed on the external switch port.

Another point to check would be the firmware and driver that the ethernet card has and if these levels are the ones recommended by VMware and/or the card manufacturer.

Although it is basic, if possible the Backup server should be connected to the same external switch where the ESXi host and the vCenter server are connected.

Have you increased the GB of the backup that is performed?

In the external ethernet switch where the ethernet card is connected where the backup and administration traffic goes, can this port be monitored to see if it is saturated during the backup? What values are displayed?

 

[hpoirier8]

Hi,

Thanks for your reply.

I will check the different points mentioned in your email and come back to you with a consolidated answer .

Thanks so much

hubert

[hpoirier8]

hi,

I got some more details regarding Backup snapshot and Network disruptions .

The application hosted on the VMs is a Real-Time application .

Based on documentation, it seems that the snapshot functionnality introduces :

- a short virtual machine freeze state called "Stun" time, which pauses the execution of the VM at a machine instruction level for up to 1 second.

so for near real-time application, it can inflict loss of ongoing transactions and sessions.

I found also this which can explain our problem we are facing :

"Long stun time while backing up virtual machine located in vVOL Datastores. (51538)"

thanks for your help.

[NanookFAS]

Hello Enthusiast.  Did you accept the link you posted as a reason for the issue you were facing?