The disk sector size and how it's managed really determine how well it will
perform under various access sizes. In your case remember, the disk sector
size is 4k, so we would hope that Linux is "doing the right thing" in coalescing
these 1k accesses into chunks which get the most out of your hard disks.

That's not our primary problem here, the primary problem here is LVM 
having trouble bringing all LVs online. These performance issues are
in addition to that. The LVM problem appears to have been addressed
by adding a root delay so that's that.

We have yet to pin down where any of them have come from. That this
problem appeared after an update is suspicious, yet iirc you've shown
that rolling kernels back and forth have no net impact on the symptoms.

If an update was really the root of all of this you should be able to
install natty to a separate set of disks, re-run these tests and see 
dramatically
different numbers for the 1K test runs.

We are now left with the actual configuration of your entire stack, the
loads generated by your VMs, and how they might be affecting your
IO subsystem.

-- 
You received this bug notification because you are a member of Ubuntu
Bugs, which is subscribed to Ubuntu.
https://bugs.launchpad.net/bugs/893450

Title:
  libvirt fails to start correctly because LVM is not ready

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/893450/+subscriptions

-- 
ubuntu-bugs mailing list
[email protected]
https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs

Reply via email to