[HPE VME] Backup crash/retry on Debian 10 VMs - production outages

[btridon]

Hi everyone,

I'm reaching out to the community because we're facing a critical issue with a client, and Veeam support hasn't gotten back to us yet despite a case open for 10 days (Case # 08134067).

Environment:
Veeam Backup & Replication 13.0.1.2067 (Windows)
Veeam VME plugin (HPE VM Essentials) 13.1.0.271
Affected VMs: Debian 10
HPE VME cluster: 2 nodes, with only 1 worker node

Symptom:
When the backup job starts, some Debian 10 VMs intermittently fail to back up properly: the task either goes into retry or crashes outright.
It's not happening on every run, it seems random, which makes it harder to pin down.
The problem is that these failures are causing production outages on the client side, which makes this urgent for us.
We haven't attempted an update yet, since we're not confident it would actually fix the issue (and we'd rather not touch the production environment without certainty).

My question:
has anyone here run into this kind of intermittent crash with VME on Debian 10 VMs, especially in a 2-node cluster with a single worker?
Could the single worker node be a contributing factor (resource contention, load during backup window, etc.)?
Do you have any leads, a known fix, or a workaround while we wait for support to respond?

Thanks in advance for any help, every bit of info matters given the urgency of the situation.

[cody.ault]

Hello. I took a look at the ticket. Based on the emails from last week from support, it seemed to indicate there was an issue talking to the NBD socket on the host without a worker. For us to process in hotadd mode, we have to have a worker on the same host as the VM you're attempting to backup. deploying a second worker to the other host would allow the job to process in hotadd mode instead of reading from the NBD. If the error is always related to reading over NBD, then we would likely need to involve HPE to investigate the service on the host other host. Hotadd mode would bypass that for the most part. We still communicate to NBD but not to read data.