Migration of volume fails

[Sridhar Venkat]

This is related to EMC VNX driver (EMCCLIFCDriver). Migration of volume during deployment fails with this error :

2014-10-24 07:37:03.648 3859 INFO cinder.volume.flows.manager.create_volume [req-04a4fad2-c685-4159-9c16-9bb085856375 0 3f2bb9e18c9542df8d6c264c810b2a63 - - -] Volume e0819649-73c2-4619-817f-cbdc7e6a7f1c: being created as source_vol with specification: {'status': u'creating', 'volume_size': 20, 'source_volstatus': u'available', 'source_volid': u'5b4855ca-33b7-45e1-85c6-99e84094d62e', 'volume_name': u'volume-boot-8202E4D_067A7DT-VNX_vscsi_npi-0a-e0819649-73c2'}
2014-10-24 07:37:03.680 3859 INFO cinder.volume.drivers.emc.emc_vnx_cli [req-04a4fad2-c685-4159-9c16-9bb085856375 0 3f2bb9e18c9542df8d6c264c810b2a63 - - -] Create snapshot: tmp-snap-e0819649-73c2-4619-817f-cbdc7e6a7f1c: volume: volume-Image_VNX_vscsi_npiv_DND_99P_capture_-5b4855ca-33b7
2014-10-24 07:37:53.783 3859 INFO cinder.volume.manager [-] Updating volume status
2014-10-24 07:38:13.656 3859 INFO cinder.volume.manager [-] Updating volume replication status.
2014-10-24 07:38:14.464 3859 WARNING cinder.exception [req-04a4fad2-c685-4159-9c16-9bb085856375 0 3f2bb9e18c9542df8d6c264c810b2a63 - - -] EMCVnxCLICmdError : ('migrate', '-start', '-source', 8122, '-dest', 77, '-rate', 'high', '-o') (Return Code: 7) (Output: [u'Error: migrate -start command failed.\\nThe destination LUN is not available for migration.\\n'])
2014-10-24 07:38:14.465 3859 WARNING cinder.volume.drivers.emc.emc_vnx_cli [req-04a4fad2-c685-4159-9c16-9bb085856375 0 3f2bb9e18c9542df8d6c264c810b2a63 - - -] Start migration failed. Message: Error: migrate -start command failed.\nThe destination LUN is not available for migration.\n
2014-10-24 07:38:26.900 3859 WARNING cinder.openstack.common.loopingcall [-] task <function _inner at 0x4a27f50> run outlasted interval by 5.22 sec
2014-10-24 07:38:34.510 3859 WARNING cinder.openstack.common.loopingcall [-] task <function _inner at 0x4a27f50> run outlasted interval by 2.61 sec
2014-10-24 07:38:41.041 3859 WARNING cinder.openstack.common.loopingcall [-] task <function _inner at 0x4a27f50> run outlasted interval by 1.53 sec
2014-10-24 07:38:47.025 3859 WARNING cinder.openstack.common.loopingcall [-] task <function _inner at 0x4a27f50> run outlasted interval by 0.98 sec
2014-10-24 07:38:53.479 3859 WARNING cinder.openstack.common.loopingcall [-] task <function _inner at 0x4a27f50> run outlasted interval by 1.45 sec
2014-10-24 07:38:53.787 3859 INFO cinder.volume.manager [-] Updating volume status

If a migration cannot be started, driver should exit before attempting to migrate. (probably due to lack of space).

We see
        self._client.migrate_lun_with_verification(source_vol_lun_id,
                                                   temp_vol_lun_id,
                                                   dest_volume_name)
        self._client.check_smp_not_attached(volume_name)

and

   @log_enter_exit
    def check_smp_not_attached(self, smp_name):
        """Ensure a snap mount point with snap become a LUN."""

        def _wait_for_sync_status():
            lun_list = ('lun', '-list', '-name', smp_name,
                        '-attachedSnapshot')
            out, rc = self.command_execute(*lun_list)
            if rc == 0:
                vol_details = out.split('\n')
                snap_name = vol_details[2].split(':')[1].strip()
            if (snap_name == 'N/A'):
                return True
            return False

        self._wait_for_a_condition(_wait_for_sync_status)

Default timeout is set to 365 days:
default_timeout = 525600

This code needs to be hardened, migration could fail due to lack of space, but it should be detected upfront before invoking migrate.

[Jeegn Chen (jeegn-chen)]

Fixed as part of https://review.openstack.org/#/c/141629/