• Welcome to OS2World OLD-STATIC-BACKUP Forum.
 

News:

This is an old OS2World backup forum for reference only. IT IS READ ONLY!!!

If you need help with OS/2 - eComStation visit http://www.os2world.com/forum

Main Menu

OS\2 Warp need to reboot if network connection drops (PWMON utility)

Started by runnindave, 2007.08.27, 23:26:00

Previous topic - Next topic

runnindave

Hello,

New to the OS\2 thing but am faced with an issue. We have an OS2 machine that connects to a local server. The OS\2 machine keeps dropping its wired network connection which makes us have to reboot it.

I found a utility called PWMON which is supposed to ping out to a remote location and if it fails to get a response it reboots the computer. I can get it to ping the local server, but if I disconnect the NIC it will announce it failed to connect forever without rebooting. It seems I need to run a web server on the box.

So can anyone think of a way to reboot an OS\2 Warp workstation if the network fails?

You guys are my last hope.

RobertM

I could... but I am not telling you how!!!

No, just kidding... give me a few (hours) and I will try to put together something for you and post it...

-Robert


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


runnindave

Hey Awesome, that would be great, unless....

there are two guys over your shoulder laughing as you wrote this... LOL.

So there is a way to ping our server, and if the ping fails ( due to network failure ) reboot. One thing I liked about the PWMON utility is that it would ping every 60 seconds and if the ping failed 3 times in a row it would reboot the machine. Too bad it doesnt work. This would be helpful if we had a temporary network outage or was replacing a switch or something like that.

Looking forward to your assistance.

Dave


Robert Deed

Wouldn't it be better to find out why the OS/2 machine keeps disconnecting from the network to begin with?  And, other then a driver not being properly loaded there should be no reason why you would need to restart the computer to get back on the network, it should be as simple as running a script which will take down the interface and start it again. 

Other the bad cables, i've never had any problems with any OS/2 servers which I have administered disconnecting.  I have had problems with the lan shares stop sharing, but that was a simple configuration problem, and since I changed to samba for my file sharing needs in OS/2 I haven't had any issues.  (needed over 2gig file support since my file server is used to archive 4.7gig dvd's)

Quote from: runnindave on 2007.08.28, 15:59:17
Hey Awesome, that would be great, unless....

there are two guys over your shoulder laughing as you wrote this... LOL.

So there is a way to ping our server, and if the ping fails ( due to network failure ) reboot. One thing I liked about the PWMON utility is that it would ping every 60 seconds and if the ping failed 3 times in a row it would reboot the machine. Too bad it doesnt work. This would be helpful if we had a temporary network outage or was replacing a switch or something like that.

Looking forward to your assistance.

Dave



runnindave

Honest question,

However, we have addressed every other NIC, TCPIP stack, Network Hardware etc detail to this point. That is where I am now. After all has been addressed, the above solution is the best solution. We cant afford to spend money on new hardware and network equipment for this problem.

The OS2 Warp is running a legacy application that we use and it is important we keep it. Its just becoming a hastle having to reboot it all the time.

Thanks for your input.

runnindave

Hey, RobertM

Have you had a chance to look into this? Is there a direction to point me in to be of some assistance?

From what I have seen, one has to program in PASCAL to create OS\2 applications? Is this true?

I am a Visual Basic guy so Pascal was something I learned back in the day.

Anyway I could port a VB app and run it on my Warp computer?

Dave

Robert Deed

You don't have to program in pascal, in fact the pascal compilers for OS/2 are about the most unsupported and outdated of any.

C/C++ helps more..

There is visualage basic for OS/2 however it is very old and wouldn't help in porting a VB app.



Quote from: runnindave on 2007.08.30, 16:59:31
Hey, RobertM

Have you had a chance to look into this? Is there a direction to point me in to be of some assistance?

From what I have seen, one has to program in PASCAL to create OS\2 applications? Is this true?

I am a Visual Basic guy so Pascal was something I learned back in the day.

Anyway I could port a VB app and run it on my Warp computer?

Dave



Robert Deed

Is the system trapping?  Do you have any logs?   Or is the system simply silently losing network connection.  When the connection is dropped are you just losing lan services or entire network io?  What is the hardware of the system.  Like I said, I would rather address the problem then simply work around it, there is a good deal of knowledge of OS/2 with various hardware available on this site, and it would help if we had all the information.  Again, a script could be made to do what you want, however if you have defective hardware the problem is most likely just going to become more frequent, and if it is a simple misconfiguration somewhere (or bad driver level) then it would make much more sense to fix that.

Quote from: runnindave on 2007.08.28, 17:23:28
Honest question,

However, we have addressed every other NIC, TCPIP stack, Network Hardware etc detail to this point. That is where I am now. After all has been addressed, the above solution is the best solution. We cant afford to spend money on new hardware and network equipment for this problem.

The OS2 Warp is running a legacy application that we use and it is important we keep it. Its just becoming a hastle having to reboot it all the time.

Thanks for your input.

runnindave

Thanks Robert

We have the same OS\2 setup on about three different computers in the test bed. For some reason, invisibly to me, they all exhibit the same behavior. They will run fine and then the IP connection just breaks off. Its something to do with software and deep inside the IP stack from what I have observed. Nothing in hardware seems to be the issue as we have used multiple cards and drivers. Its hard

Not being an expert in OS\2 I was not able to track down what was going on or what was causing the break. I had read somewhere that this is a known issue with OS\2 Warp ( one comment in some blog FWIW ) and there was no solution.

So my quick solution is to simply reboot when the machine drops its connection. I just have no clue on how to script this.

I have a REBOOT.exe file that works fine. If there was some script that could ping out every minute, and if the ping fails after the third ping, call the reboot.exe file.

that would be ideal for us.

Thanks for your time.


RobertM

Sorry Dave,

I got a little busy with some work projects, but am looking into it.

Can you please provide me with the following info...

  • What version of OS/2 are you using on the machine (and test machines) in question?
  • Have you uprgraded or installed fixes for any of the following... TCPIP, REXX, networking in general?

OS/2 (as of WSeB) had a "bug" in the original TCP/IP stack (which was fixed in later releases such as WSeB CP2 and CP2PF) that would cause the stack to lock and no longer route or respond to TCP/IP traffic if the machine received a certain type of attack/packets (I dont remember the specifics of the cause, but this might be your issue).

I run WSeB CP2PF (and have run the earlier releases) and once I switched to CP2 and CP2PF (or applied the TCPIP fixes to WSeB original) never ran into the problem you are describing again. I've run multiple websites from the same box - which handles ooodles of traffic (lots too! ;) ) and not experienced any issues since applying the fix or switching to a later release of WSeB.

I *had* a reboot tool I used to use (but can't find it - will rewrite it for you)  that I used for that and other situations... what I did was assign a machine an IP like this... XX.XX.XX.100, and then an alias on the same card for XX.XX.XX.150 - my second machine was XX.XX.XX.101 with no alias. If the second machine saw no response (via pinging the first) on IP XX.XX.XX.150, it would alias itself to that address and assume the functions of the first machine. The first machine, when rebooted would tell the second machine "I'm up... un-alias that IP" and then check to see when the 2nd machine dropped the alias. It would then alias itself to the .150 address and resume handling traffic.

I should be able to re-write that script so that it checks any other machine or location on the net, and failing a response, initiates a reboot. This can be handled a few different ways... either a simple ping utility that times out properly (some dont seem to, and just hang waiting for a ping response) or using something like wGet or Curl to retrieve a small page from someplace with a timeout - and if that timeout occurs, initiate a reboot.

So that brings up the next question... do you have a webserver on another machine that has a small page (or that you can create a small page on) that the script can attempt to access? If so, I can have a script written pretty quickly... it would use some custom stuff in REXX and the GetHTTP library (or wGet) from Hobbes.

But that brings up one final question... what version of REXX are you using? (ObjectREXX or Classic REXX)? ObjectRexx eventually crashes the REXX.DLL after an app is running for a long time using Stem Arrays - there is a better version on Hobbes that doesn't crash as quickly - but you may not need it depending on what version of Warp you are using (WSeB CP2PF and eCS 1.X DO show this problem - but unless you are using lots of REXX scripts that create a lot of Stem Arrays, you should have that problem... and if you are running CRexx and not ORexx, you wont run into the problem at all).

-Robert


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


runnindave

Hey, Glad you are still with me....

I will get anything you need.

I am using OS\2 Warp 4.0 (Internal Revision 14.062_w4)

There have been no upgrades to the existing Network stack. Here is why, this was an image created by a group that are no longer around for support.

Your Fail over Reboot tool sounds incredible. Not what we need here, but I know if you can make that fail over process work, I feel really confident that you can help me.

We do have a website that it can ping out to. Its an intranet web site.

Hmm, Rexx version...Not sure how to get it.

I REALLY appreciate your time on this BTW, I hope with a little direction, I can make this as easy and less time consuming for you.

Dave


RobertM

Hey Dave,

OK... for the REXX thing (you definitely HAVE Rexx... it's just a matter of which one you are using - though on Warp 4 I dont think it matters)... all you need to do is open a command window and type the following:

Quoteswitchrx [ENTER]

It will come up with a few lines of info... hit "N" [ENTER] when it asks you to type yes or no... but read the screen... it will say something like this:

QuoteThis program will now replace the current Classic REXX interpreter of OS/2
with the Object REXX interpreter.

Enter 'Y' to continue, any other input will cancel this procedure.

The first line is the relevant one... tell me what that says...


Now, as for the machine and setup itself... you can grab OS/2 Warp 4 disks pretty cheap when you can find them on eBay... you are apparently running standard Warp 4.

There are a bunch of fixes for it, including a new TCP/IP stack... they are mostly all rolled into one package. You can grab Warp 4 FixPack 15 various places on the Internet and installing it is as easy as following the instructions. Then you can upgrade the TCP/IP stack as well - also pretty easily. I can help you through that process if you like - it will probably alleviate all of your issues... and as a backup I can write something to reboot the machine (and even do fail/fallover to another if you have a second machine that can take over). Keep in mind that if you try using a second machine as a backup, you have to be mirroring the data to it so that both machines keep current... without knowing more about your setup, the method of doing that would vary.

Let me know, and I will come up with something... I have a test script in the works. I'll adjust it as necessary.


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


runnindave

Hey,

OK I followed your instructions and we got the same identical first lines.

QuoteThis program will now replace the current Classic REXX interpreter of OS/2
with the Object REXX interpreter.

Enter 'Y' to continue, any other input will cancel this procedure.


Update, apparently a group of IBM experts could not narrow down this problem. All of the lastest IP stacks were upgraded to the latest version. The final statement was "Just reboot."

So we are back to this script. Hopefully we can automate the reboot and its sounds like you are close. I will say I am impressed with OS\2 and am not sure why it is not more common as an useful OS. Was it the Microsoft machine that hurt it? It seems like it may have been ahead of its time.

Thanks for your assistance with this RobertM.

Let me know if there is anything else you need.

Dave

RobertM

Hi Dave,

I will work on completing the script... though if you and your team are willing, I'd also like to try to resolve this problem. I've sent you a private message with my direct email address with requests for more information if you want to try to resolve this.

Basically, the problem you are having shouldn't be happening. I've run many servers on different OS's and found OS/2's stack to be one of the most reliable - even under horrendously heavy load. I just completed another load test on my Netfinity machine and shoved 100 simultaneous connections down it's throat totaling 140,000 connections when the test was done... neither the machine nor the TCP/IP stack cared in the least bit. All of the servers my company installs are running OS/2 or eCS, and after gigabytes worth of TCP/IP traffic and months of uptime, the only thing I have to go out to do at the clients' are clean dust from the machines. Some of my older machines were as "underpowered" as PII 300's that handled over 200,000 TCP/IP requests a day without issue.

So... basically what I am saying is (as far as TCP/IP is concerned) you should probably never have to reboot your machine. Ever.

So, what I propose is the following... I'll get the script done as a short term fix, and if you are willing, help you track down and resolve the issue so the script later becomes a non-issue.

-Robert


|
|
Kirk's 5 Year Mission Continues at:
Star Trek New Voyages
|
|


runnindave

Awesome!

As you said, I would prefer to implement the script first as a quick fix so we don't have to monitor these OS\2 boxes after hours. But it sounds like you are up for a challenge and it may be. I will setup a time to work with you on the fix and maybe we can solve the riddle.

Keep in mind that we are not able to provide consulting fees other than a six pack and some cheetoes. LOL.

Thanks again

Dave