Hi. I use some seriously antiquated hardware to run my picture and video servers for my site Scene.com.

Each server runs Win95 osr2 with MS Personal Web Server 1.0a. One is a Pentium Pro 180 with 32MB ram. The other is a 486dx33 with 24MB ram. They work great most of the time. :mrgreen:

But for some reason, PWS will sometimes stop working. If you watch the system montior graphs, the CPU usage will be at 100% and the server will stop responding to http requests. However, it will still respond to local network requests and ftp requests. It seems to do this more during times of heavy loads (such as yesterday when they were serving about 23 users a minute (just a guess from the host server log since I don't have the logs on these servers). During heavy loads, the servers may not even stay running for 30 minutes without needing to be rebooted. :cry:

I've been reading and researching solutions to my deliemma. I've read about memory leaks, and issues with win95/98 winsocks having memory leaks. At this point, I'm trying to determine if I need to change just PWS or the platform that it is running on (I'm considering NT since it will run decently well on the same hardware, but this is a last resort as it will be a pain since I have even less time than money).

Here's some links that I've gathered:





If anyone has an experience or feedback, please feel free to post. It's an interesting problem from a computer standpoint, but a pain in the butt from a business standpoint. :lol: Any help is appreciated.

Dani AI

Generated

Quick summary and a practical triage path based on the symptoms described by and the hardware comment from .

The fact that FTP and local networking continue to work while HTTP stops strongly points to the web server (or its HTTP listener) getting stuck or running out of a resource rather than a full network outage. Focus the investigation on what the HTTP process is doing at the moment of failure and how sockets/connections behave.

  • Reproduce and capture the failure quickly. When it happens, use Task Manager (or any process viewer) to note which process shows high CPU, memory, and thread count.
  • From another machine, test the HTTP listener with a raw connect: telnet <server-ip> 80 then send GET / HTTP/1.0. If the TCP connect succeeds but no HTTP response arrives, the server threads are hung; if the connection fails, the listener may have crashed or unbound the port.
  • Run netstat -a on the server to look for many TIME_WAIT or CLOSE_WAIT entries (signs of socket exhaustion). If you see many stale TCP states, the service could be running out of available connections.
  • Capture packets from a separate machine or mirrored switch port (not on the old box itself) to see whether connections are accepted, reset, or dropped — this avoids overloading the antique hardware with capture tools.
  • Check/enable PWS logging and rotate logs; examine access patterns just before a failure for a particular URL, script, or large upload that coincides with the crash.

Short-term mitigations

  • Disable dynamic modules/CGI that you can, and disable keep-alives or reduce their timeout so sockets aren’t held open.
  • Move caching/static content to a small front-end box (a cheap secondary PC or small Linux box) as a reverse proxy to absorb spikes.
  • Test with native Windows networking (as you planned) to rule out Lantastic or third-party network layers.

If the server still fails under modest load, plan an upgrade path: either move the site to a slightly more modern and supported OS or put a lightweight front-end proxy in front of the Win95 boxes.

Recommended Answers

All 2 Replies

Sigh.....your hardware is extremely antiquated. The reason why your system hangs is that your CPU is being used up completely(100% as you said), and no CPU time is available to service the http requests. I'd suggest STRONGLY that you get a new computer stat.

Another problem could be your internet connection. See if that isn't running at a low speed. But I suspect your hardware. It may be at it's last leg....

I think you mis-understood the cpu 100% issue. The cpu usage is under 20% most of the time, but when PWS stops working, the usage will be 100%. That's the way I know to check it.

The Internet connections are two 3mb/384k cable modems. They'll put out almost 100kbyte/sec together. I know they're fine.

Since the time of this post, I have tried Webweaver (www.brswebweaver.com) and it worked for an even shorter time before needing rebooting. There may be an issue with Lantastic being used for the network connection. I'll be trying the built-in Win networking soon and see if that doesn't work better.

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.