tests: ceph_test_rados_api_watch_notify: test timeout using rados_wat…

[tchaikov]

…ch3()

objecter resends the watch request upon tcp reconnecting, and OSD
resets the watch timeout when handling the resent watch request, so,
with a small "ms tcp read timeout", the timeout test always fails. in
this change, we use rados_watch3() which supports the "timeout" option,
so we can pass a timeout smaller than "ms tcp read timeout", and hence
the test can pass.

Fixes: http://tracker.ceph.com/issues/19312
Signed-off-by: Kefu Chai kchai@redhat.com

[tchaikov]

/home/jenkins-build/build/workspace/ceph-pull-requests/qa/workunits/cephtool/test.sh:1891: test_mon_cephdf_commands:  ceph osd pool create cephdf_for_test 32 32 replicated
2017-03-21 06:43:24.529379 7f01d9399700 -1 WARNING: all dangerous and experimental features are enabled.
2017-03-21 06:43:24.542077 7f01d9399700 -1 WARNING: all dangerous and experimental features are enabled.
Error EINVAL: crushtool check failed with -22: crushtool: timed out (5 sec)
/home/jenkins-build/build/workspace/ceph-pull-requests/qa/workunits/cephtool/test.sh:1: test_mon_cephdf_commands:  rm -fr /tmp/cephtool.rXN

The following tests FAILED:
	  5 - cephtool-test-mon.sh (Failed)

see https://jenkins.ceph.com/job/ceph-pull-requests/20521/consoleFull#428417753c5f5a223-5c83-43f2-bbe3-2257412b27db

retest this please