Multiple Telnet Processes Using Excessive CPU

[Glen Cook]

We have a V2500/6-way application server accessed via telnet. When our DNS server bounced people could connect to the app server but would hang at various points of login, (DNS is back up at this point). Glance showed many telnet processes using high CPU, (pegged at 100%), and it took 30-40 minutes to clear out. Although there were around 50-75 users trying to access the system at the same time it would not normally be a problem. There were no errors in any logs and no other symptoms. Why would these telnet processes use so much CPU and take so long to clear out?

[rick jones · Accepted solution]

They were probably spinning on something or other. Perhaps failed DNS calls. In the hopefully unlikely event that it happens again, you might use tusc (ftp://ftp.cup.hp.com/dist/networking/tools/) to take a system call trace of some of the sessions.

A less detailed view of the system calls could be found via Glance.

[Glen Cook]

Rick,

Thanks for the input. nslookup -swtrace showed that DNS was good during this time which is why I was puzzled it would take another 30 minutes to clear up. I'll take a look at tusc in case we have a repeat performance. Thanks again!

[Judy Traynor]

You could have an application running on top of the telnets that is actually causing the problem. I had a similar situation about a year ago with the Dec 2002 HP11 patch bundle, so I would recommend that you have telnet nfs, arp patches at least as recent as March of this year.

[T. M. Louah]

Consider patching telnetd, the following patches :
PHNE_22159: HANG MEMORY_LEAK
PHNE_21952: HANG MEMORY_LEAK
The latest telnetd patch is PHNE_26096
Good luck
!T+