Detect stale conns when ping times out 03/71203/5
authorJosh <jhershbe@redhat.com>
Mon, 23 Apr 2018 10:11:54 +0000 (13:11 +0300)
committerVishal Thapar <vthapar@redhat.com>
Wed, 23 May 2018 04:42:34 +0000 (04:42 +0000)
commitd60780fbd8701e2aea2c12d18ac2b987d2e6def0
tree993600cff67bc2968118755a6397d34fd1fd2576
parent2a2e875637c7ce3a42843bc5b2ecb813e79a6214
Detect stale conns when ping times out

When an ovsdb client connects StalePassiveConnectionService
checks to see if there are any previous connections to that
same client. If there are, they are pinged to determine
whether or not they are still actually connected. The
callback that handles the ping results does not fire
properly in the event of a timeout and the onFailure method
does not call
OvsdbConnectionService.notifyListenerForPassiveConnection
for the new connection. This results in the new connection
not being being reported "up the stack."

Note: easiest way to reproduce this is to "unplug" (destroy
in libvirt) a VM running the OVS and then start it.

Issue: NETVIRT-1178
Change-Id: Ie3494c46719954f862d8edb6e8921a752dcd30ea
Signed-off-by: Josh <jhershbe@redhat.com>
library/impl/src/main/java/org/opendaylight/ovsdb/lib/impl/StalePassiveConnectionService.java