13 Failures of domjobabort

[cevich]

On 11/05/2013 03:48 PM, Dongsheng Yang wrote:

On 11/04/2013 07:45 PM, John Ferlan wrote:

  13 domjobabort

Another similar to previous, but in this case the shutdown happens in
the middle of the test for some unknown reason... It'll take some more
research...

I did not reproduce this issue
The last commit on my branch is
750874c

Is it possible that your laptop
works with too heavy load and then VM did not start? Just my guess.
Could you run this test once again to verify this error? thanx.

Environment:

• Lenovo T530, 4 CPUs (dual-core + HT), 8G ram
• Latest hand-build of libvirt (almost 1.1.4)
• Running as root, SELinux in permissive mode
• Default bridge mode networking
• Default JeOS 19 x86_64 guest

[cevich]

Typically what I find in cases such as this is some other test that
failed results in leaving the environment in a shall we say unusable
state. What I do to determine that is run the test separately. If it
succeeds, then I go backwards through the list of tests to find which
test may be leaving things in a indeterminate state. I believe in this
case, it's the domif_setlink_getlink failures which have repurcussions.
Before applying your change I tried the following:

• If I just run domjobabort, it succeeds
• If I run domiflist,domiftune,domjobabort - all succeed
• When I included domif_setlink_getlink in the list I had 3 failures in
  domiflist, each of the domiftune tests took quite a bit longer (250s vs.
  a norm of 8s), and 13 FAIL's for domjobabort (each also had 250s runs
  vs. 12-13s in a good run). The whole run took 3H20M to complete.

After applying your fix, resetting myself to top of trunk, and
re-running the 3rd bullet resulted in the hang mentioned above.

The delay I saw in bullet 3 makes sense since the typical/default
timeout is 240s for wait_for_login(). Probably fixes a number of the
other issues I saw...

So maybe this is related to #1009 also?