[rpc-test] make RpcPendingConnectionsMetric more stable
I noticed that on very fast machines the RpcPendingConnectionsMetric
scenario would rarely fail (~1 time per 500 runs) because the
sock_diag() netlink facility seemingly reported stale data, i.e.
reported one connection was pending while the only attempted connection
had already been accepted by the server and even closed by the client.
With this test no flakiness has been observed in more than 50K runs
of the scenario.
Change-Id: If8e66c471452b1e04c84cf2d5c979578c287b4fa
Reviewed-on: http://gerrit.cloudera.org:8080/21251
Tested-by: Alexey Serbin <alexey@apache.org>
Reviewed-by: Mahesh Reddy <mreddy@cloudera.com>
Reviewed-by: Abhishek Chennaka <achennaka@cloudera.com>
1 file changed
