PVC provisioning with insufficient storage

Root Cause Analysis

This case need to be tested on both RWO/RWX volumes

  1. Create a PVC with size larger than 8589934591 GiB.
    • Deployment keep in pending status, RWO/RWX volume will keep in a create -> delete loop.
  2. Create a PVC with size <= 8589934591 GiB, but greater than the actual available space size.
    • RWO/RWX volume will be created, and the associated PV for this volume will have annotation “longhorn.io/volume-scheduling-error”: “insufficient storage” in it.
    • We can observe “Scheduling Failure” and “Replica Scheduling Failure” error messages on the Longhorn UI with the following details
      • Scheduling Failure
      • Replica Scheduling Failure
      • Error Message: insufficient storage
  3. Create a PVC with size < the actual available space sizeļ¼ŒResize the PVC to a not schedulable size
    • After resize PVC to a not schedulable size, both RWO/RWX were still in scheduling status.

We can modify/use https://raw.githubusercontent.com/longhorn/longhorn/master/examples/rwx/rwx-nginx-deployment.yaml to deploy RWO/RWX PVC for this test

[Edit]