Intermittent cinder volume timeout with IBM Storwize

Issue

While using IBM Storwize as a backend for cinder service and when image_volume_cache_enabled is set as True, the following conditions were observed :

• Cinder configuration file with image_volume_cache_enabled=True, configuration can be found under the file /etc/cinder.conf on the controller.

  
  [ibm]
  image_volume_cache_enabled=True
  image_volume_cache_max_count=50
  image_volume_cache_max_size_gb=1000
  san_ip=x.x.x.x
  san_login=admin
  san_private_key=/etc/cinder/ibm.key
  storwize_svc_volpool_name=xxxx
  volume_backend_name=ibm
  volume_driver=cinder.volume.drivers.ibm.storwize_svc.storwize_svc_fc.StorwizeSVCFCDriver
  backend_host=hostgroup
  rpc_response_timeout = 1800
  
  

• Fast operations:

  • Cloning volume with the same size
  • Using volume cache where new volume has the same size
  • Extending a volume created from an image
  • Manually extending a cloned volume if the backend has already completed the FlashCopy operation

• Slow operations:

  • Cloning volume to a bigger volume
  • Using volume cache where new volume has a bigger size
  • Manually extending a cloned volume if the backend is still doing the FlashCopy