Description of problem:
When the daemon of xend stops ,the cron task who is /etc/cron.hourly/mcelog.cron can't read /sys/hypervisor/uuid.It leads to D status of "cat /sys/hypervisor/uuid " which results in high load average
Version-Release number of selected component (if applicable):
In RHEL5.4 for 64 bits x86_64 xen kernel.
In xen kernel,make the daemon of xend stop.
Steps to Reproduce:
1.#service xend stop
2.#top (see the load average by hours)
3.#ps aux |grep "^D" (there are lots of "D" status process)
See the result of 'ps' ,we will find there are a lot of "D" state process.Theses process are waiting for cat /sys/hypervisor/uuid.And the load average become more and more high.
When the daemon of xend stops, the process can read /sys/hypervisor/uuid.The load average is not high by hours.
In RHEL 5.4 before,we find the bug for Reading /sys/hypervisor/uuid in Dom0 hangs if XenStoreD isn't running.The bugzilla website is https://bugzilla.redhat.com/show_bug.cgi?id=225203
Interesting. I can still read /sys/hypervisor/uuid after stopping xend. Are you sure it's not exactly the same issue as in bug 225203? That is, is xenstored running when you try to read uuid?
Changing component to kernel-xen to decide if we want to fix this bug this time. Kernel tries to read xenstore and blocks forever if xenstored is not running. There is even a patch in bug 225203. We could also reopen that bug and mark this one as a dup of it. I'll let kernel-xen developers decide...
Is this not the same as BZ 522827 ?
Yes I would agree. Duplicate?
*** This bug has been marked as a duplicate of bug 522827 ***
Clearing out old flags for reporting purposes.