Using the NVM SSD of the Hetzner VM with Longhorn

[JakobStadlhuber]

Is it possible to use the available NVM SSD with Longhorn for applications that need faster speed like DBs where Hetzner Block Storage is probably the wrong solution?

[M4t7e]

Yes, this is actually the default behavior. By default, Longhorn uses /var/lib/longhorn on the host to store data. The /var directory is the dedicated EPHEMERAL volume provided by Talos, which uses the local NVMe SSD (system disk).

[JakobStadlhuber]

That's amazing and Longhorn keep track of scaling nodes or deleted nodes? I do not have to worry that if a node is going down or is deleted that my files are away? what if I delete or rotate all my nodes, example switching all nodes to a higher server tier?

[M4t7e]

Longhorn is a distributed block storage system that stores multiple replicas of your data across different nodes. If a node is permanently lost, such as when it is down or deleted, Longhorn automatically rebuilds the affected replicas on healthy nodes to maintain data redundancy and availability. However, if all nodes hosting a Longhorn volume are deleted at the same time and there are no backups, the data will be lost.

Tier upgrades typically do not involve deleting nodes, but they do require a reboot. This can make data temporarily unavailable if all related nodes are down at the same time. I recommend learning how Longhorn works so you can plan your node pool configuration for these operational tasks. See: https://longhorn.io/docs/

[JakobStadlhuber]

Thank you

[JakobStadlhuber]

One last question:

What about scaling nodes will Longhorn deploy there Volumes there too? Can i for example flag/tain it that only certain fixed nodes space is used by Longhorn?

[M4t7e]

Yes, and you can also control which nodes are allowed to host volumes. See the documentation:

• https://longhorn.io/kb/tip-only-use-storage-on-a-set-of-nodes/
• https://longhorn.io/docs/1.10.0/references/settings/#create-default-disk-on-labeled-nodes