Problem with Streams Memory

[ad73]

Hi:

I am running SCO Openserver 5.0.5 and all was running fine until we added a CISCO SWITCH and some DIGI portservers to add a wireless component to our network. Since adding these components my SCO Server loses TCP/IP Connectivity every couple of days. I did a netstat -m and it seems to be running out of Streams memory. The Network shuts down everytime the "MAXIMUM STREAMS MEMORY USED" and "STREAMS MEMORY IN USE" approaches the "TOTAL CONFIGURED STREAMS MEMORY". Does anyone know how to fix this? Also, can anyone explain to me WHY this is suddenly happening. We have been using the same system for the Last three years and I have never even had to so much as reboot the server since day 1. Any help would b appreciated. Incidentally, I have both a WINDOWS TERM SERVER and this SCO SERVER attatched to the same Cisco Switch and the WINDOWS TERMINAL SERVER has not gone down even once. Thanks in advance. By the way, I'm still sort of a UNIX newbie.

ad73

 

[ppalanca]

Sounds like you're having resource issues.

To boost memory allocated for streams....must go to tune hardware/kernel....choose 12/streams....try doubling the amount for [nstrpages]...believe default is 500?? We've set this as high as 2000.

You'll need to relink/reboot.

Also...you might want to add supplement oss497C which addresses network...stuff.

Check on /usr/adm/syslog and /usr/adm/messages....these log files may show that the server is getting bombarded with network messages.
Digi's can cause issues with "unreachable" ports or respawn errors when both terminal and modem ports are enabled [ie:ttya01 and ttyA01]. This will pipe so many messages to your log files that resources get used up to the point of bringing server down. Serial port issues will bring server down eventually. Streams memory would be an after-effect since resource/memory is severely hampered.

 

[ad73]

Thanks ppalanca:

I think you're right that I'll have to Tune the STREAMS MEMORY SETTINGS, but for clarification, the Server does not DROP only the networking components. The Server itself and all of its local tty's (eg tty01,02,03 etc) are active and fully functional. Also, I have not enabled the ports on the Digi's as modem ports, only as Terminal ports.

Any other ideas...AND THANKS FOR THE QUICK RESPONSE.

By the way, I am looking for a file on my server called gated.conf in the /etc directory as I read a help article saying that this might have some bearing on my situation. I searched everywhere for the file but can't find it in /etc.

 

[ppalanca]

Digiportserver II? Using Realport? Set your ports to [prn] and realport to 771.

So the memory allocated for network services gets overloaded and you eventually lose it. Increasing memory probably will only delay its eventual demise.

Since the additions to your network...the server is receiving packets of information that it may not be able to resolve. That...or there's a process running on network that's using the allocated memory for streams [ie: a program running on a PC that hooks up to server via TCP]. Examine the "wireless component". Is it accessing the server in some way?
Had a similar issue where streams went down due to client running a Window's program that ran a process which flooded TCP with "packets" that server could not resolve. It was a program that ran a search on server for a simple file lookup. Easily enough done ON server....but outside the server is another story. The process used streams like a cache file...till it would crap out.
You might define the wireless IPs in /etc/hosts. Or research /etc/routed to resolve incoming packets from outside server [man routed]. [gated.conf] is a dead end.
Examine /etc/inetd.conf...is bootps active? Pound it out.
Finally...use netstat commands to examine TCP...maybe even delete TCP protocol through netconfig and reinstall NIC card.

 

[GorillaDF]

I had a similar problem that was driving me nuts! Turned out bootpd was giving errors to syslog almost 1/second! Commented out, restart inetd, and instantly streams memory in use dropped from 1200 to 331!

 

[markwilkinson]

Did you ever get to the bottom of the streams problem on SCO Openserver 5.0, as I appear to be having similar issues.? By that I mean that the network will run for a few hours and then crash. On checking the streams resources (with netstat -m) I discover that 1 or more streams classes has a large number of failures. Rebooting the system will bring it back to life. I have tried doubling the amount for [nstrpages] but this at best appears This has now been happening for the last 4 days - before this period I never had any unplanned reboots - the system just ran without any problems for weeks at a time.

Any help would be most appreciated.


Mark

 

[ad73]

Hey Mark.

Yeah, we did finally figure out what was going on. It seems that SCO v5.0.5 needs to have a service patch installed after you make these changes to the server (ie in our case we installed Digi portservers with Real port Driver). Anyways, its been a while so I don't have the exact patch information, but if you go to the SCO website and search under ALL proucts for "Out of Streams Resources" you will find the patch that may be right for you.

Hope this helps a little.

ad73

 

[markwilkinson]

You are quite right there is a patch available on SCO's web site. (Support Level Supplement, SLS OSS497, It seems you must install Release Supplement 505A first if you have not already done so.) The notes for SLS OSS497 do mention that it should address issues related to the depletion of streams memory. Unfortunatly, having applied this software to my system it has 'nt prevented my network from stopping after 15 or 20 hours.

I plan to take the server off our live network over the weekend and just have it plugged into a test hub or switch. I'm kinda hoping that this will help me estiblish if it's a problem with the server (such as a faulty NIC), or a problem on our live network (such as a virus - although we have AV software installed).

I will report back once I've got to the bottom of this - hopefully sooner rather than later!

Many Thanks

Mark

 

[markwilkinson]

As mentioned I took the server off the live network and placed it on a "Test Network" during a quiet period. The problem then went away. At this point it was looking like a virus or other faulty network device. Last Friday we discovered a couple of PC's on the live network that had not been updated with the AV software. These were upgraded right away to the latest version of SOPHOS, along with all the other PC's.

Although to the best of my knowledge we have had no virus alerts after this point the Unix box has now stopped running out of streams resources and has run for over 3 days without a reboot - it would barly run for 20 hours before. On checking the "maximum streams memory used" (with netstat -m) I can see that this value is now stable. Before it was gradually creeping up to the point where it would bring down the network.

This is the best explination I can give to how we have solved our problem.

 

[ad73]

Hi Mark:

Great to hear that things are working again at your end. Thanks for keeping me posted.

ad73