Note: This bug is displayed in read-only format because the product is no longer active in Red Hat Bugzilla.

Bug 1113989

Summary: Nova live migration with LIO seems to be broken
Product: Red Hat OpenStack Reporter: Lukas Bezdicka <lbezdick>
Component: openstack-cinderAssignee: Eric Harney <eharney>
Status: CLOSED CURRENTRELEASE QA Contact: nlevinki <nlevinki>
Severity: high Docs Contact:
Priority: high    
Version: unspecifiedCC: eharney, nlevinki, sclewis, scohen, sgordon, sgotliv, yeylon
Target Milestone: z4Keywords: TestOnly, ZStream
Target Release: 5.0 (RHEL 7)   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: openstack-cinder-2014.1.3-2.el7ost Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2015-04-14 18:47:25 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
cinder log output none

Description Lukas Bezdicka 2014-06-27 12:09:20 UTC
Created attachment 912792 [details]
cinder log output

Description of problem:
Issue 1)
We have node A and B where A is controller and compute B is compute.
User starts instance with volume.
Instance gets scheduled on B as it has little more resources but cinder creates volume on controller.
cinder-rtstool create /dev/cinder-volumes/volume-62e7b56a-ccf7-4a9f-b987-f136c0a8ab2b iqn.2010-10.org.openstack:volume-62e7b56a-ccf7-4a9f-b987-f136c0a8ab2b QRTDqf4gRFC6Nddh8Pmh 7faS6KYkQwde2wMk2Ars

Volume target is created BUT also initiator for node is created even though it does not have to:
initiator_name = None ; name_file = '/etc/iscsi/initiatorname.iscsi' ; acl_new = rtslib.NodeACL(tpg_new, initiator_name, mode='create')

So we just got initiator created for node A with credentials of node B.
Cinder now creates initioator for node B.

cinder-rtstool add-initiator iqn.2010-10.org.openstack:volume-62e7b56a-ccf7-4a9f-b987-f136c0a8ab2b QRTDqf4gRFC6Nddh8Pmh 7faS6KYkQwde2wMk2Ars iqn.1994-05.com.redhat:5ede4ee65025

Now user wants to livemigrate and cinder creates initiator for node A.

cinder-rtstool add-initiator iqn.2010-10.org.openstack:volume-62e7b56a-ccf7-4a9f-b987-f136c0a8ab2b QLL8NXdFkty76daetceL VgSYZL4n2oB7NFxH7FTR iqn.1994-05.com.redhat:78f884f357c5

Oh wait that initiator exists and it has some password set, we'll just keep that.

This fixes this issue 1)
https://github.com/openstack/cinder/commit/37472be084358717ec2a046147e42dc248258ee9

Issue 2)
Let's say we applied patch for issue 1).
Cinder creates target for volume and creates initiator for node B.
We try to livemigrate and cinder creates initiator for node A.
After succesfully connecting to iscsi on node A and moving the instance from node B it just calls cinder-rtstool delete and drops the whole target of running node. See attachement.

Comment 4 Eric Harney 2014-08-27 14:00:48 UTC
This upstream bug probably explains this.

Comment 8 Sergey Gotliv 2014-12-04 14:31:53 UTC
Eric, please update status of that bug.

Comment 9 Eric Harney 2014-12-04 18:42:30 UTC
Change was already pulled in via 2014.1.3 rebase.