1

We moved our retail software to new hardware (we passed from an IBM x3200 with SATA 7200rpm disks to an IBM x3200 with SAS 15000rpm disks) but after a few days with no problem at all we keep getting

Aug 29 08:39:50 server01 kernel: [64680.439253] retail-soft[1089]: segfault at 104 ip ? 081106ea sp bfd1b7c0 error 4 in retail-soft[8048000+18b000]

in /var/log/messages and the software just hangs or gets closed.

We changed the RAM assuming that was the problem but it keeps happening.

What else can we check?

We have installed the latest stable release of Debian. We also discarded a programming failure because the same software is running in other servers with no problems at all.

|improve this question
  • 1
    What is retail-soft? From what I can understand it's a service. Is that correct? – Karlson Aug 31 '12 at 14:47
  • More like a bug in your commercial software maybe – daisy Sep 15 '12 at 10:44
1

Why do you assume that the problem is in hardware? Could it be that the software reaches a condition under which it produces a Segmentation fault?

If you really want to know whether or not you have a problem with memory I would suggest looking for ECC errors in you /var/adm/messages logs or in the remote console if you have one. You can also run Memtest86 which on Ubuntu is installed by default and is available from the boot menu, I am not sure about Debian but could be the same.

You cannot discard programming failure under any circumstances until you have proven that there is no other cause for failure. Software running on other servers would only be an indicator if:

  1. Software performs the exact same function
  2. Software services same number of requests
  3. Software services the exact same requests
  4. Configuration of the software is 100% identical
  5. Configuration of OS outside the hardware drivers is 100% identical

Outside of that I would start with the core analysis to see if the software crashes in the exact same place every time.

|improve this answer
  • Thanks Karlson, retail-soft it's the sale system we use, we ran memtest and it passed OK all of the tests. Now I checked free -m and said total = 2015, when it's 8GB physically – Gonzalo Aug 31 '12 at 17:37
  • 2
    ok, that indicates two things: 1. the kernel isn't detecting all available RAM when it boots, 2. the retail-soft program isn't checking whether a memory allocation succeeded, it's just assuming it did and seg-faulting when it tries to use it. Your question is now "how do i get the kernel to recognise all the memory?". You can add a mem= option to the kernel boot command line in /etc/default/grub – cas Sep 1 '12 at 2:29
  • 2
    BTW, if you want more than 4GB you need to either run a 64-bit Debian, or a 32-bit debian with a 64-bit kernel or the 32-bit + PAE extensions kernel. A PAE kernel will recognise all 8GB but individual processes can only use a maximum of 4GB each. – cas Sep 1 '12 at 2:31
  • @Gonzalo If the memtest pass all the tests you can be reasonably sure that the issue is not physical. – Karlson Sep 4 '12 at 15:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.