[Sunhelp] TCP Problems?

[Ann Waters]

Hi Listers,

Machine: SunOS sodns5 5.6 Generic_105181-15 sun4u sparc SUNW,Ultra-5_10

I've got an odd (to me) situation occurring that I haven't a clue as to how to troubleshoot.  One of my DNS/DHCP servers has "connections" to the outside world as follows:

1.  System health monitoring via SMTP and sysedge.
2.  Two TCP connections to a NetID database for DNS/DHCP.

Every once in a while, the sysedge system health report will show that the DNS/DHCP services are down.  If I should catch this when it actually happens, I attempt to telnet to the machine but get "connection refused".  (If I already had an open telnet session, it winds up getting closed.)  Once the sysedge system health report shows that the DNS/DHCP services are back up, I can telnet in to try to check things out after the fact.  My /var/adm/messages has logged some errors for the "down time" indicating that the two TCP connections to the NetID database were lost.  The DNS/DHCP processes were still active from their original start date, so they really weren't down as reported by the sysedge software.  I set top up to run continouosly with 30 second updates just to see if there was some sort of process overload, etc.  The top output looks like the system is just humming along.

This seems to have started happening regularly on Sept. 16, with occurrences now at 2-3 times a day (sometimes more).  The switch that the box is connected to supposedly tests out okay.  The machine hosting the NetID database is having no problems with any of my other DNS/DHCP servers that are also connected via TCP.  So, it looks like the problem is w/in this one box.

Does anyone have any ideas/suggestions on tools/processes I can use to try to narrow down just what is going on with TCP?  It would have to be something that I could start running via cron (or at startup) and just leave, since there is no definate time, etc. as to when this happens.

Thanks in advance for any/all help!

Ann Waters
Kaiser Permanente
Silver Spring, MD