inetd daemon hanging

[VINU]

Hi All,

The inetd daemon on one of our server is hanging intermittendly.Once the inetd daemon is restarted the problem get solved temporarily.

The OS is 10.20 and the model of the server is 9000/847/G30/H30.

I founf the following error message related to it,
"Networking memory for fragment reassembly is restricted to 8470528 bytes"

Request Your valuable inputs regarding this.

Regards
Vinu

[Patrick Wallek]

I would first try installing the latest 'inetd cumulative' patch (PHNE_20747) for 10.20, which is from June 2000.

http://www1.itrc.hp.com/service/patch/patchDetail.do?BC=patch.breadcrumb.main|patch.breadcrumb.search|&patchid=PHNE_20747&context=hpux:800:10:20

I would also go ahead and install the very latest (also from 2000) General Release patch bundle for 10.20 so you are as up to date as possible on patches. You can get it here:

http://software.hp.com/portal/swdepot/displayProductInfo.do?productNumber=B3782XSW1

[Dave Olker]

Hi Vinu,

The "network memory" message you're referring to is a normal, informational message. It is not a warning or an error - it merely tells how much memory is configured for IP fragment reassembly. I really doubt that has anything to do with inetd hanging.

In addition to installing the latest inetd patches (per Patrick's suggestion), I suggest you start inetd in full-blown debug logging mode, also known as enhanced connection logging mode. To do this, simply start inetd with the "-b" option. This option is not documented in the man pages (at least it wasn't the last time I checked) but with this logging enabled, if inetd hangs again you will hopefully have a good idea what inetd was in the process of doing when it stopped. The logging goes to the /var/adm/syslog/syslog.log file.

Hope this helps,

Dave

[Sundar_7]

"Networking memory for fragment reassembly is restricted to 8470528 bytes"

I believe the kernel parameter netmemmax is the one that controls the amount of memory that can be used by the networking subsystem for fragmenting the packets.

I am not sure if this has got something to do with inetd hanging. But if the problem seem to persistent, set netmemmax to unlimited and see if that helps.

[rick jones]

It has been an exceedingly long time since I've worked with 10.20, but if my recollection is correct, if there is indeed insufficient memory for IP fragment reassembly, the netstat statistics would have lost of entries in the IP stats for fragments dropped "due to dup or out of space"

And I concurr that it is very unlikely to be related to a hung inetd.