Yes, absolutely. If you configure Longhorn with volume replication, the Talos worker on Proxmox will seamlessly keep serving disk reads and writes without missing a beat, even if the Linux Mint laptop completely goes down or disconnects.

How Longhorn HA Works Across Your Setup

When you create a volume in Longhorn (for example, a 10 GB persistent volume for a database), Longhorn creates a Volume Engine on the node where the pod is running, and Replicas on the underlying disks across your cluster:

┌││││└─────P---──r──oTPR──xaoe──mldp──ool──xs(i──Ac──VWpa──Eop──r1──Nku──oes(──driP──enr──1gi──m──da──ir──sy──k──)l──o──c──a──l────c──o──p──y──)─────────────────────────┐││││┘◄─────────1─7─.─6T─h─Gu─bn─pd─se─r─Nb─eo─tl─wt─o─r4─k───────────►┌││││└─────L--──i──nTR──uae──xlp──ol──Msi──ic──nWa──to──r2──Nk──oe(──drS──ey──2n──c──i──n──g────b──l──o──c──k────d──a──t──a──)─────────────────────────┐││││┘
  1. Synchronous Writes over Thunderbolt: Every time your application writes data, Longhorn writes it locally to Replica 1 (on Proxmox) and sends it over the 17.6 Gbps link to Replica 2 (on Linux Mint).
  2. Read Efficiency: Reads are served locally from Proxmox with zero network latency.

What Happens When Linux Mint Goes Down?

Here is the exact step-by-step sequence when the Linux Mint laptop disconnects or reboots:

  1. Heartbeat Loss: Longhorn detects that worker 2 (Linux Mint) is unreachable.
  2. Replica Degraded State: Longhorn marks Replica 2 as Degraded or Offline, but keeps the Volume Engine running.
  3. Zero Downtime for Applications: Any pod running on the Proxmox node continues reading and writing to Replica 1 without any interruption or read/write errors.
  4. Automatic Healing upon Reconnection: Once the Linux Mint laptop powers back on and rejoins the cluster, Longhorn automatically re-syncs all changed blocks over Thunderbolt to bring Replica 2 back up to date.

Key Requirements to Test This HA Setup

To ensure this high availability test works smoothly, keep these 3 settings in mind:

1. Set Replica Count = 2

In your Longhorn StorageClass or volume configuration, set numberOfReplicas: 2. This forces Longhorn to keep one copy on Proxmox and one copy on Linux Mint.

2. Configure Node Anti-Affinity

Longhorn automatically enforces soft/hard anti-affinity, meaning it will refuse to put both replicas on the same Talos node if two nodes are available.

3. Adjust Longhorn Offline Timeout Settings

By default, Kubernetes takes ~5 minutes to evict pods from an unreachable node (pod-eviction-timeout). However, for the storage volume layer, Longhorn handles node loss almost instantaneously (within seconds).