Comment 24 for bug 1373513
Revision history for this message
| John Griffith (john-griffith) wrote | #24 |
© 2004
Canonical Ltd.
•
Terms of use
•
Data privacy
•
Contact Launchpad Support
•
Blog
•
Careers
•
System status
•
0e1f616
(Get the code!)
Finally had a chance to look at this more closely since I can easily reproduce it with Rally. It turns out that the issue is we're running vgscan on all disks on the system. What happens though is that some of those disks are a result of a cinder create / attach, and we hit issues where the attachment is removed and deleted during the vgscan causing the scan operation to hang and if we're lucky timeout.
We can fix this by modifying lvm.conf to ONLY query devices that we specify as being used by cinders VG, this has risks though incase somebody wants to use the system for other purposes. Turns out though you can create your own lvm.conf and set it via an env variable.
Running a number of tests iterations with Rally (which produces the original issue every time), this works and cleans up the issue. I'll just need to figure out the best way to deal with env vars in Cinder commands.