Create iscsi volume from image fails when use_multipath_for_image_xfer=True

[Lucio Seki]

Cinder fails to create an iSCSI volume from image when multipath is enabled for image transfer, and no targets were discovered yet.

Steps to reproduce:
- prepare a host with no iSCSI targets discovered yet;
- in cinder.conf, set use_multipath_for_image_xfer=True;
- try to create an iSCSI volume from an image.

Expected result: an iSCSI volume to be created successfully.

Actual result: TargetPortalNotFound exception is raised by os_brick/initiator/connectors/iscsi.py.

[Lucio Seki (lseki)]

Analyzing the logs, I observed that the execution sequence is as follows:

connect_volume -> _connect_multipath_volume -> _get_ips_iqns_luns -> _discover_iscsi_portals -> _get_node_startup_values -> _run_iscsiadm_bare -> ProcessExecutionError: Unexpected error while running command.

Where ProcessExecutionError says:

ProcessExecutionError: Unexpected error while running command.
Command: iscsiadm -m node --op show -p 10.250.118.22:3260
Exit code: 21
Stdout: u''
Stderr: u'iscsiadm: No records found\n'

and is followed by another ProcessExecutionError:

Command: iscsiadm -m discoverydb -o show -P 1
Exit code: 21
Stdout: u'SENDTARGETS:\nNo targets found.\niSNS:\nNo targets found.\nSTATIC:\nNo targets found.\nFIRMWARE:\nNo targets found.\n'
Stderr: u''

This exception is catched by _get_ips_iqns_luns, who raises as exception.TargetPortalNotFound.

This error occurs because _discover_iscsi_portals attempts to discover targtes, but first it tries to get a list of already discovered "old_node_startups", and fails if no targets were discovered yet.

The bug can be solved by making _get_node_startup_values accept exit_code=21 from _run_iscsiadm_bare, indicating that there are no old node startup values at all.

[Gorka Eguileor (gorka)]

This was fixed when we fixed https://bugs.launchpad.net/os-brick/+bug/1756206 with review https://review.opendev.org/#/c/554704/