ping interval in HP-UX

[Shabahang]

i have a block lost problem on 10gR2 rac cluster.
Cluster wait is abnormally high on one of the 2 nodes.
i wanna check the interconnect connectivity, by a ping flood. but cant put a ping with a small Interval.
command (ping -I host )
accepts only integer .
please help me with my test.
1- any other idea with connectivity test
2- how to recude the interval value

[Steven E. Protter · Accepted solution]

Shalom,

Linux ping might be a better tool here.

It has a -f flag that really floods the network.

I'm not sure ping with hpux can do what you need it to do.

SEP

[rick jones]

Statistically speaking, just using a normal one second interval and waiting for two minutes should give similar results.

Have you already looked at netstat and lanadmin statistics on the cluster node?

You could I suppose download netperf (www.netperf.org) and then ./configure --enable-histogram and run a ten second TCP_RR test between the suspect node and some other system with a verbosity of 2:

netperf -H -t TCP_RR -v 2

and then look at the resulting round-trip time histogram. Those that are >> the average RTT will likely have been retransmitted one or more times.

[Shabahang]

Tnx ,