Applicable versions

All Longhorn versions.

Symptoms

When creating a Pod with Longhorn Volume as PVC, the Pod cannot start.

When checking for error message using kubectl describe <pod>, the following message is shown:

Warning  FailedAttachVolume  4m29s (x3 over 5m33s)  attachdetach-controller     AttachVolume.Attach failed for volume "pvc-xxx" : rpc error: code = Internal desc = Bad response statusCode [500]. Status [500 Internal Server Error]. Body: [message=unable to attach volume pvc-xxx to node-xxx: volume pvc-xxx not scheduled, code=Server Error, detail=] from [http://longhorn-backend:9500/v1/volumes/pvc-xxx?action=attach]

Noticed the message returned by Longhorn in the error above:

unable to attach volume pvc-xxx to node-xxx: volume pvc-xxx not scheduled

Details

This is caused by Longhorn cannot find enough spaces on different nodes to store the data for the volume, which result in the volume scheduling failure.

Most common reason

For Longhorn v1.0.x, the default Longhorn installation has following settings:

1. Node Level Soft Anti-affinity: false.
2. The default StorageClass longhorn's Replica count is set to 3.

That means Longhorn will always try to allocate enough space on three different nodes for three replicas.

If this requirement cannot be satisfied, e.g. due to there are less than 3 nodes in the cluster, the volume scheduling will fail.

Solution

If this is the case, you can:

1. either set Node Level Soft Anti-affinity to true.
2. or, create a new StorageClass with replica count set to 1 or 2.
3. or, adding more nodes to your cluster.

Other reasons

See scheduling section in the Longhorn doc for a detail description of the scheduling policy.

Related information

Starting Longhorn v1.1.0, we will introduce a new setting Allow Volume Creation With Degraded Availability(true by default) to help with the use case on smaller cluster.

See https://github.com/longhorn/longhorn/issues/1701 for details.