Description of problem: With fresh OpenShift Origin on Fedora 19 installation as of today, running oo-diagnostics seems to be stuck in the test_node_profiles_districts_from_broker step. The oo-diagnostics process blocks after INFO: broker application cache permissions appear fine INFO: running: test_node_profiles_districts_from_broker INFO: checking node profiles via MCollective at recvfrom(14, "=)\201\203\0\1\0\0\0\1\0\0\6broker\7example\3com\3"..., 2048, 0, {sa_family=AF_INET, sin_port=htons(53), sin_addr=inet_addr("10.16.96.113")}, [16]) = 116 poll([{fd=14, events=POLLIN}], 1, 4999) = 1 ([{fd=14, revents=POLLIN}]) ioctl(14, FIONREAD, [116]) = 0 recvfrom(14, "\263\261\201\203\0\1\0\0\0\1\0\0\6broker\7example\3com\3"..., 1932, 0, {sa_family=AF_INET, sin_port=htons(53), sin_addr=inet_addr("10.16.96.113")}, [16]) = 116 close(14) = 0 uname({sys="Linux", node="cloud-qe-2.idm.lab.bos.redhat.com", ...}) = 0 futex(0x63746d4, FUTEX_WAKE_OP_PRIVATE, 1, 1, 0x63746d0, {FUTEX_OP_SET, 0, FUTEX_OP_CMP_GT, 1}) = 1 futex(0x6374750, FUTEX_WAKE_PRIVATE, 1) = 1 futex(0x1a5bf64, FUTEX_WAKE_OP_PRIVATE, 1, 1, 0x1a5bf60, {FUTEX_OP_SET, 0, FUTEX_OP_CMP_GT, 1}) = 1 futex(0x1a5bf30, FUTEX_WAKE_PRIVATE, 1) = 1 futex(0x1a5c634, FUTEX_WAIT_BITSET_PRIVATE, 47, {6941, 514163964}, ffffffff for multiple seconds (minutes?) and cycles. When I kill the process, it says FAIL: rescue in block in run_tests error running test_node_profiles_districts_from_broker: #<OpenShift::NodeException: Could not connect to ActiveMQ Server: SIGTERM> INFO: running: test_broker_accept_scripts INFO: running oo-accept-broker I tried to service activemq restart but it did not help. There is nothing interesting (== pointing to error) in /var/log/activemq/*. Version-Release number of selected component (if applicable): OpenShift Origin nightly on Fedora 19 as of today. How reproducible: Deterministic, I have seen this on multiple installations. And I did not see it yesterday. Steps to Reproduce: 1. Do OpenShift Origin nightly on Fedora 19 installation. 2. Run oo-diagnostics -v -w 1 Actual results: The process gets "stuck" at or around the test_node_profiles_districts_from_broker step. Expected results: See it go through the steps without getting stuck anywhere, like shown in bug 974497. Additional info:
Actually, I can see the same behaviour on installation with separate node + the other machine having all the other services as well.