Filepro conversion part 2

Yesterday I began the story of a conversion from Windows to Linux. At first, everything seemed to be going well. Then some strange things started happening.

I was accessing this through a Windows Terminal Server and then sshing over. They had called me because I forgot to transfer a minor file, so I connected and then attempted a "su". It wouldn't take my password. I hadn't yet set myself up for sudo, so I was stuck. I assumed I had misremembered the password, though that seemed odd, as I had been using it just the day before. But I had left ALT-F1 logged in as root, so I had them run "passwd" to reset it and we continued.

The next day things were going well enough that they wanted to do some printed reports. I hadn't set up printers yet, so I connected and again had the same problem: I couln't su. I asked if they had changed anything, nope. Very, very strange. I had them reset it again. This time I added myself to the sudoers file. But I had to do it manually: visudo wouldn't run. It dumped core, and I was starting to feel very uncertain about the whole machine. That was escalated a moment later when I tried to "cat" a file and got a segmentation fault. The "cat" worked.. but it gave that error.. very weird. I immediately thought of the original installation problems I had seen with Filepro: their install script had been failing with segmentation faults.. what the heck was gong on here? I decided a site visit was in order.

When I got there I did a shift pageup on the F1 console and saw that nobody had messed with the password other than the two resets I had requested. I also confirmed that "cat" and now several other commands were issuing seg faults. Most were not working at all; it was odd that cat was at least "working". I decided to shut the machine down.

That was the end of life for this machine. It came back up, but only "sort of". Networking wouldn't start (segmentation faults, of course) and there were many other failures. This had to be very low level: corruption in a library or the kernel itself.

So that's where we are right now. They didn't have another machine handy to repeat the install, but I happen to have another box coming to me that was slated for another project. There's no rush on that, so I'll temporarily steal that for this test. I'll have to do everything all over again, but most of that is just unattended computer time, so it's not too bad. With luck, we can get back to testing on Monday.

And we did.. that new box has been running flawlessly ever since. Reports that used to run overnight now take seconds.. and zero network locking problems. Fast, reliable.. good move!

Psst - Wanna work for yourself?

Comments


Fri Jul 28 14:43:14 2006: Subject:   BigDumbDinosaur
Scripts that segfault? Veeeerrrrrrrryyyyyyyy ssstttrrraaannngggeee, as Arte Johnson's character would have said on the old "Laugh-In".



Fri Jul 28 14:56:46 2006: Subject:   TonyLawrence
Well, binaries that segfault. I suppose I could have traced it down with trace or just looking at ldd for common libraries, but why bother? There's too much risk..



They said the box always ran find under Windows. Of course that's meaningless..

Sat Jul 29 02:51:55 2006: Subject:   drag
Is that machine using ECC memory?



Sounds like RAM problems to me. That seems most likely.
Like the bad portion is in the kernel memory area that is causing file system issues. Something like the change to the password file is happily sitting in cache, but it is not writing to the file correctly. Since the Windows kernel and subsystem stuff is much larger then Linux kernel and such then it makes sense that Linux has a decent chance of running better on a flaky machine.

Personally I would write the box off for the time being, buy some el'cheapo box for temporary fix or conscript a desktop box for a stand in and bring the harddrive over. If it makes the problem go away then you know the hardware is the problem. If it's a modern Linux system then everything should autodetect and be setup correctly.

Then if that is the problem take the machine back or run memtest86 and see if you can figure out what exactly the problem is.

Sat Jul 29 11:41:08 2006: Subject:   TonyLawrence
Yeah, that was my feeling. And that's what I did..



Wed Aug 2 00:24:37 2006: Subject:   bakerd
I supported a filepro application running on linux for eight years. We moved to a new oracle-based software application in years six and seven, but for historical data reasons, we had to keep the filepro application intact. As a part of the project, I installed a new oracle database server. Since the filepro server was now six years old, I wanted to move the application to the new server and retire the old one. The old server ran Red Hat 7.3. The new server was running Red Hat ES 4. When I attempted to move the app to the new server, the filepro binary seg faulted every time. Just for "fun" I tried filepro on Red Hat 8 and 9 as well, each with the same problem. I don't recall the actual conflict I discovered, but there was definately a library conflict between the newer versions of Red Hat and filepro. I didn't have any relationship with filepro directly to discuss the issue and, ultimately, I wrote an application to convert all of the filepro data into Microsoft Access to preserve the data and retire the old server.



Wed Aug 2 10:11:44 2006: Subject:   TonyLawrence
You wouldn't have been able to just "move" it without relicensing or at least setting the new machine's MAC address to match the old. Unlicensed Filepro does segfault (or at least it does on Linux - you'd think they'd put up a "Invalid License" message instead).



Thu Dec 21 20:38:01 2006: Subject:   TonyLawrence
New box has been running beautifully for months.. and so much faster than the old Windows box.. reports that used to take "too long" run in seconds..

---

Lone-Tar Backup and Disaster Recovery
for Linux and Unix

---

Fri Dec 22 15:31:09 2006: Subject:   BigDumbDinosaur
New box has been running beautifully for months...

That's 'cuz your client is no longer using a kid's coaster wagon to haul rock and gravel to the job site.



Add your comments

Why no "Digg this!" etc.?

Enter your email address for automatic notification of new posts here
(be sure to whitelist 'feedburner.com' if you use spam filtering)

Control Spam and Viruses

Run your own mailserver

Have you tried Searching this site?

Unix/Linux/Mac OS X support by phone, email or on-site: Support Rates

This is a Unix/Linux resource website. It contains technical articles about Unix, Linux and general computing related subjects, opinion, news, help files, how-to's, tutorials and more. We appreciate comments and article submissions.

Publishing your articles here