topic Re: node crashed due to buffer space not available in Operating System - HP-UX

node crashed due to buffer space not available

mor_1 — Mon, 02 Aug 2004 00:35:46 GMT

Hi,
After a package had been sccussesfully started on a running node, the node just crashed in about 15 minutes!
I checked the OLDsyslog.log and it has the following messages:
Aug 2 11:29:38 SD2 cmcld: GS connection to 192.1.1.1 is responding
Aug 2 11:45:38 SD2 cmcld: accept returned: No buffer space available
Aug 2 11:45:38 SD2 cmcld: Retrying accept due to a transient problem: No buffer space available.

swapinfo shows the following messages:
dev 1048576 0 1048576 0% 0 - 1 /dev/vg00/lvol2
dev 11534336 0 11534336 0% 0 - 1 /dev/vg00/lvswap
dev 6144000 0 6144000 0% 0 - 1 /dev/vg00/lvswap2

total physical memory is 18841600 Kbytes

I dont know what causes the problem, and I need your help.

Re: node crashed due to buffer space not available

Steve Steel — Mon, 02 Aug 2004 02:00:43 GMT

Hi

Probably a patch missing for your serviceguard . use itrc to check your patch level

Patch Name: PHSS_30028

Patch Description: s700_800 11.X MC/ServiceGuard and SG-OPS Edition A.11.14

Fixes a very similar problem

If you have PHSS_29561 it has a problem and should be replaced by the 30028 asap

Steve Steel

Re: node crashed due to buffer space not available

Jeroen Peereboom — Mon, 02 Aug 2004 02:02:00 GMT

L.S.

you can use the Search More options at the top of this page to search the HP-UX forum for buffer space available. I found http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=190143

but there are more threads on this subject.

JP.

Re: node crashed due to buffer space not available

Mei Jiao — Mon, 02 Aug 2004 02:23:34 GMT

Latest patch for that MC/ServiceGuard patch is this:
PHSS_31015 s700_800 11.X MC/ServiceGuard and SG-OPS Edition A.11.14

Re: node crashed due to buffer space not available

mor_1 — Mon, 02 Aug 2004 02:33:29 GMT

Thanks guys!
I'll try the patches.

Re: node crashed due to buffer space not available

Jose Mosquera — Mon, 02 Aug 2004 03:33:43 GMT

Hi,

Have you done a massive LAN port scanning?

When this happend a telnet reconnection storm occurs being able to exhaust the available buffer space for this purpose. Then a TOC (Transfer Of Control) is generated by MC/ServiceGuard. A suggested action to avoid this could be:

1.- Install last ARPA Transport comulative patch (PHNE_29473) and dependencies (reboot required).

2.- Ensure that MC/ServiceGuard NODE_TIMEOUT parameter (/etc/cmcluster/cluster.conf) have a value among 5-8 to minimize the TOC risk. If you need increase this value you must stop the pkg, execute the cmapplyconf command and restart the pkg.

Rgds