* We create the volume and target in Cinder
* Nova grabs the volume info and starts building the instance
* Nova issues the discovery, and actually succesfully logs in to the target
* Nova attempts to actually make the attach and map to /dev/disk-by-path
The last step is what actually never seems to complete in the allowed period of time with the given retries.
I don't have a good theory here as I haven't been able to reproduce the error, but the only thing that I can suggest at this point is that we either increase the retry factor from 2 to 3 to give us almost 30 seconds before the last retry, or the same thing could be accomplished by changing the default for CONF.num_iscsi_retries
Looking at this and here's my take:
* We create the volume and target in Cinder
* Nova grabs the volume info and starts building the instance
* Nova issues the discovery, and actually succesfully logs in to the target
* Nova attempts to actually make the attach and map to /dev/disk-by-path
The last step is what actually never seems to complete in the allowed period of time with the given retries.
I don't have a good theory here as I haven't been able to reproduce the error, but the only thing that I can suggest at this point is that we either increase the retry factor from 2 to 3 to give us almost 30 seconds before the last retry, or the same thing could be accomplished by changing the default for CONF.num_ iscsi_retries