3PAR: concurrent volume creation causes multiple driver issues

Bug #1658888 reported by Vijay Ladani
14
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Cinder
New
Undecided
Unassigned

Bug Description

Casue of failure:
-----------------
When a user wants to perform parallel cinder volume creation operation, 3par driver ends the execution up with below different errors in different test scenarios.

1. iSCSI login failure "Could not login to any iSCSI portal" during Concurrent volume creation from image fails for 3par iSCSI backend iscsi login error
2. VolumeBackendAPIException: Bad or unexpected response from the storage volume backend API: Driver initialize connection failed (error: Not found (HTTP 404) 17 - host does not exist).
3. ImageCopyFailure: Failed to copy image to volume: Bad or unexpected response from the storage volume backend API: Unable to fetch connection information from backend: Conflict (HTTP 409) 73 - host WWN/iSCSI name already used by another host

Proposed fix:
-------------
Applying synchronization on initialze_connection and terminate_connection methods of 3pariscsi and 3parfc drivers addresses all above mentioned issues.

Performance analysis:
---------------------
Performance data captured for a specific test to measure the impact of synchronization. Test case is to create 10 OR 20 concurrent image volumes with different scenarios and observe the results.

Here is the performance result for all different test scenarios...

Average time taken by single test operation without synchronization:

1. Create and delete 10 concurrent 1g volumes with cirros image(image size is approx 25MB) - 20 to 22 seconds
2. Create and delete 10 concurrent 40g volumes with ubuntu image(image size is approx 600MB) - 92 to 95 seconds
3. Create and delete 20 concurrent 40g volumes with ubuntu image(image size is approx 600MB) - 179 to 185 seconds
4. Create and delete 10 concurrent 80g volumes with ubuntu image(image size is approx 600MB) - 93 to 95 seconds

Average time taken by single test operation after synchronization:

1. Create and delete 10 concurrent 1g volumes with cirros image(image size is approx 25MB) - 29 to 31 seconds
2. Create and delete 10 concurrent 40g volumes with ubuntu image(image size is approx 600MB) - 88 to 90 seconds
3. Create and delete 20 concurrent 40g volumes with ubuntu image(image size is approx 600MB) - 173 to 175 seconds
4. Create and delete 10 concurrent 80g volumes with ubuntu image(image size is approx 600MB) - 87 to 89 seconds

From the above results, we conclude that there is a very less impact on performance rather improved performance as the volume size go higher after applying synchronization.

Note: performance data captured and displayed here is through rally test framework.

Vijay Ladani (vladani)
affects: rally → cinder
Changed in cinder:
assignee: nobody → Vijay Ladani (vladani)
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix proposed to cinder (master)

Fix proposed to branch: master
Review: https://review.openstack.org/424454

Changed in cinder:
assignee: Vijay Ladani (vladani) → Vijay Ladani (vijay-ladani)
status: New → In Progress
Revision history for this message
Sean McGinnis (sean-mcginnis) wrote : Bug Assignee Expired

Unassigning due to no activity for > 6 months.

Changed in cinder:
assignee: Vijay Ladani (vijay-ladani) → nobody
Changed in cinder:
status: In Progress → New
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Change abandoned on cinder (master)

Change abandoned by Sean McGinnis (<email address hidden>) on branch: master
Review: https://review.openstack.org/424454
Reason: No updates in quite awhile. Feel free to restore and update if this is still needed.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.