3com 3812/3824 switch

[rwd1]

Hello forum.

Just signed up, first post

I am a sw developer and not a network admin so pleas bear with me. . .

I have an issue which I believe is caused by a 3812 or 3824 switch going quiet for a period of about a minute and a half.

The issue is that there are redundant servers trying to communicate with each other on connected sockets through these switches.  When communication is lost (the sockets stay connected ). . . just no data xmit & receive. If the outage is long enough then each server thinks it has lost its partner and changes state to operate on its own. . . Then data transfers resume and things are sorted out (not without a certain amount of drama) by the system & its operators.

This really isn't a significant issue in the end . . .  just a pain to have to explain. Any ideas as to what may be happening which would cause this quiet period and how to perhaps monitor via SNMP.

Thanks 

Rich

[Vince-Whirlwind]

Can you please provide sanitised switch configs?

Make sure the interfaces these servers are attached to have Description on them, as well as any other relevant interfaces, eg uplinks.

Also, a copy of the log file off both switches.

Setup a continuous ping between two other hosts that are one on each of the two switches and pipe the results to a file. When the servers stop communicating, does the ping between the two other hosts show anything?

Setup a continuous ping between the two servers that are having the issue. same deal.

[rwd1]

Thanks for the interest.

How does one generate a "sanitised switch config"?

Remember I am a  sw developer and am venturing far afield here. . .

In so far as the continuous pings, I believe that is what the software in the apps are essentially doing.

Specifically one server is attempting to access a database table in the other to

read the state that server thinks it is in.

In another case a computer is sending a "heart-beat" message and then waiting for a reply.

Both of these events stop cold for about 1 & 1/2 minutes. . .  then resume. . . but the damage (change of operational state) has been done.

Could this be happening, if the switch is running the Spanning Tree Check? 

Rich

[Vince-Whirlwind]

A sanitised config is a config that can't help a hacker. Remove passwords, SNMP strings, usernames, public IP addresses.

The purpose of running continuous pings is so you can see for yourself what the state of the network connectivity is, rather 

If you are dubious about spanning-tree then you would need to provide your network topology in addition to the switch configs and logs.

[rwd1]

Just checked and Spanning Tree is dis-abled. . . So can't be that.