HP-UX 11.11 hang - TCP Wrapper bug?

[Gordon Morrison_1]

I've been playing with a couple of test systems with a view to beefing up security (a learning experience!) I ran Security Patch Check & installed the recommended patches, then installed SSH and played with it - everything seemed fine. Then I installed TCP Wrappers, and after I got over the initial configuration problems, everything seemed to be working - but then the server (a K370 running 11.11) started hanging every night, and yesterday it hung every hour or so. I couldn't get a crash dump from the console with ^B and tc (though rs works, strangely enough). Even hitting the TOC button on the back gets no response.
I restored the original inetd.conf last night and rebooted, and I thought that had fixed it: It stayed up all night, but then it hung again late this morning.
Today as an experiment I installed TCP Wrappers on the other test box (SSH already installed) - a K200 running 11.00, and it has been stable all day.

Has anyone run across (and fixed) this sort of behaviour before?

Alternatively, can anyone suggest any way of getting a crash dump out of this K370 when it is hung. The only way I have found to get it back is either to power-cycle it, or reset it from the console - either way I get no crash dump.

Thanks

[Steven E. Protter]

Few queries:

1) Are there core dumps?
2) What does dmesg /var/adm/syslog/syslog.log say.


Suggestions:
1) Run a complete hardware check with cstm mstm or xstm
2) Possbily temporarily pull tcp wrappers out of production on this server for a few days. If the problem stops, you have your culprit. If not, its another issue.

This could be caused by something as simple as another box attempting to grab your ip address. If you are on the public internet, spammers sometimes try and grab an ip not assigned to send a little spam.

Just one of many possibilities. My gut tells me hardware though.

SEP

[Gordon Morrison_1]

As luck would have it, it crashed last night and it did produce a dump, though I can't be sure it was caused by the same problem.
When it hangs, I can't get a dump as everything is just locked up and I have to powercycle it. syslog just stops.
This box is not on the internet, just our own LAN.
I took TCP wrappers out of the equation (I just restored the original inetd.conf - I didn't uninstall the software) yesterday morning, and it hung again in the afternoon.
stm threw up a couple of interesting points in the CPU activity logs:

-- Information Tool Activity Log for CPU on path 32 --

Log creation time:
Wed Oct 27 10:22:29 2004

Wed Oct 27 10:22:29 2004: Information tool (cpu) starting on path (32).

Wed Oct 27 10:22:29 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa0000).

Wed Oct 27 10:22:29 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa2000).

Wed Oct 27 10:22:29 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa0000).

Wed Oct 27 10:22:29 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa0000).

Wed Oct 27 10:22:29 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa2000).

Wed Oct 27 10:22:29 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa2000).

Wed Oct 27 10:22:29 2004: Tool completed with exit_status SUCCESSFUL (0)
indicating tool completed without errors.



-- Information Tool Activity Log for CPU on path 34 --

Log creation time:
Wed Oct 27 10:23:12 2004

Wed Oct 27 10:23:12 2004: Information tool (cpu) starting on path (34).

Wed Oct 27 10:23:12 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa0000).

Wed Oct 27 10:23:12 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa2000).

Wed Oct 27 10:23:12 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa0000).

Wed Oct 27 10:23:12 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa0000).

Wed Oct 27 10:23:12 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa2000).

Wed Oct 27 10:23:12 2004: Internal error : Failed to get the configuration
status of the CPU. HPA of processor in question is
(0xfffffffffffa2000).

Wed Oct 27 10:23:13 2004: Tool completed with exit_status SUCCESSFUL (0)
indicating tool completed without errors.


...Though I have to admit I don't know how serious this is: it reports errors, then says the tool completed without errors?

[Gordon Morrison_1]

It was a flaky CPU, now replaced