It seems some Bug reports are hard to reproduce because reporters (sometimes even developers) use different cluster configuration than CSIT tests.
Possible sources of this situation are outdated wiki pages, outdated parts of User Guide and other ReadTheDocs documentation, outdated templates and scripts laying around in Integration/Test repository, and perhaps even Youtube videos, testing VMs and similar. We should actively search for such sources of outdated information and delete them, or at least attach a big warning. The only recommended way to configure cluster (Beryllium, Boron, Carbon) is configure-cluster.sh script (to be moved to Controller [0] soon). Of course, some alternative configurations should work as well, but some are actively discouraged, as they lead to known failure modes; due to Akka limitations or Controller requirements. Typical example is auto-down. Do not use it unless you are willing to deal with cluster splitting to parts (in some isolation scenarios). So please, if you encounter such an outdated source, reply to this thread (possibly after fixing the source right away). Vratko. P.S.: Similar care should be taken when describing test scenarios. Isolating all traffic is different from isolating all TCP traffic, which is different from isolating only ports 2550 and 2551. Killing Karaf is not the same as stopping Karaf. Restarted Karaf can have persisted data wiped, or not. [0] https://bugs.opendaylight.org/show_bug.cgi?id=7569
_______________________________________________ controller-dev mailing list controller-dev@lists.opendaylight.org https://lists.opendaylight.org/mailman/listinfo/controller-dev
