summaryrefslogtreecommitdiff
path: root/include
diff options
context:
space:
mode:
authorJean-Sébastien Pédron <jean-sebastien@rabbitmq.com>2017-06-08 16:50:02 +0200
committerJean-Sébastien Pédron <jean-sebastien@rabbitmq.com>2017-06-13 12:12:47 +0200
commit0bd9368048704de5464605896a6c09befd173221 (patch)
tree34c1ae5055e79566c519cd7cf78ec386c2cb7ab7 /include
parent085820636c5c52c0b61900ae1597e9f2ecd740b4 (diff)
downloadrabbitmq-server-git-0bd9368048704de5464605896a6c09befd173221.tar.gz
partitions_SUITE: Try to improve partial_false_positive
The testcase simulates a small delay between the times B sees A down and C sees A down too. Because A is effectively down for both, B's rabbit_node_monitor shouldn't see a partial partition. When B sees A down first, it asks C about A. In the scenario the testcase simulates, C tries to contact A, fails and reports to B that A is down. However, because of the way we simulate that delay, sometimes, C still sees A when B asks about it, because the link between A and C isn't blocked yet. When this happens, B decides there is a partial partition. Therefore, the testcase fails because it tests that this shouldn't happen. To try to be closer to a real-world situation (where A would be really down for B and C), we suspend C's rabbit_node_monitor until we block the link between A and C. Thus B still asks C about A, but C only processes the request when the link is indeed down. [#146911969]
Diffstat (limited to 'include')
0 files changed, 0 insertions, 0 deletions