'Runaway' Nodemgr revisited

Hi Mark,

We've experienced two problems with nodemgr running away with memory:

1. All the memory on the box gets chewed up.

2. Just the memory of the nodemgr partition suddenly runs out of

memory.

Peggy Adrian gave me these patches that solved the first problem:

{...snip...}

Also, here are the patches forte is recommending:

1) 103582-11 through 103582-15

2) Change the 'TCP Qlen for Half-Open Connects' setting to

zero. Make the following change to your Solaris machines:

# ndd -set /dev/tcp tcp_conn_req_max_q0 0

There is no need to reboot the Solaris system after the change. You

*will* need to restart the Forte Environment Mgr (and repos server).

3) Contact SunMicro to get SunOS 5.5.1 Patch-ID# 103582-16

built May/22/98, or the SunOS 5.6.x equivalent.

For the second problem, after working with support for three weeks, we

finally found out that our connected environments were somehow corrupt.

Both of our two connected environments thought the other was the super

environment and that they had no subs. When a partition went down in the

first environment it would check its super environment before auto-starting

a new partition. The second environment would not have a copy of the

partition either so would check with its super environment (which happened

to be the first one). The two connected environments started shouting at

each other until thousands of threads were started thus consuming all the

available partition memory. In this case we simply rebuilt the environments

and all is well again.

Hope this helps.

Scott Ostlund

Carlson Wagonlit Travel

Sostlundcarlson.com <<a href=

"mailto:Sostlundcarlson.com">mailto:Sostlundcarlson.com</a>>

612-594-2561

--Original Message--

From:Mark Carruthers [<a href=

"mailto:MarkCafox.com">mailto:MarkCafox.com</a>]

Sent:Monday, January 03, 2000 2:18 PM

To: kamranaminyahoo.com

Subject:(forte-users) 'Runaway' Nodemgr revisited

Back in March and April there were several emails on the topic of runaway

nodemgr processes on Solaris 2.6 but I can't find any mention of a fix

(other than using ulimit to kill the processes when they use too much

memory).

We've just encountered the problem and it seems to affect other Forte

processes as well as the nodemgr process.

Did anybody ever find a solution?

TIA,

Mark Carruthers

20th Century Fox

--

For the archives, go to: <a href=

"http://lists.sageit.com/forte-users">http://lists.sageit.com/forte-users< ;/a> and use

the login: forte and the password: archive. To unsubscribe, send in a new

email the word: 'Unsubscribe' to: forte-users-requestlists.sageit.com

at 2007-6-29 9:16:12 >