Thread

1. constant crashing hardware issue and thank you TAKE AWAY

   jack <jack4pg@a7q.com> — 2024-04-17T13:06:31Z

   I discovered that one of the memory sticks in the machine was damaged.
   Running memtest86 on the machine generated many RAM errors.
   This was causing the strange bi-polar errors in postgresql.
   
   The hardware technician explained that he sees this often and that there is no one cause for such problems.
   
   As I am not a hardware specialist, I never thought that RAM could cause such problems.
   I always assumed that the OS (ubuntu or windows) would advise me if there was ever an issue with memory.
   
   TAKE AWAY:
   As a result of this I will be checking the RAM on all my machines once a month or the moment a machine starts to act strange.
   
   Thanks again to all who helped with this issue.

2. Re: constant crashing hardware issue and thank you TAKE AWAY

   Madalin Ignisca <hi@madalin.me> — 2024-04-17T13:08:33Z

   That kind of support for “damaged ram” you have it with ECC memory on CPU’s that support it.
   
   XEON cpus for example.
   
   > On 17 Apr 2024, at 15:06, jack <jack4pg@a7q.com> wrote:
   > 
   > uld advise me if there was ever an issue with me
   
   

3. Re: constant crashing hardware issue and thank you TAKE AWAY

   Justin Clift <justin@postgresql.org> — 2024-04-17T16:19:17Z

   On 2024-04-17 23:06, jack wrote:
   <snip>
   > As a result of this I will be checking the RAM on all my machines once
   > a month or the moment a machine starts to act strange.
   
   Once a month is overkill, and unlikely to be useful. :)
   
   With server or enterprise grade hardware, it'll support "ECC" memory.
   
   That has extra memory chips + supporting circuity on the memory board
   so it can detect + correct most errors which happen without them causing
   problems.
   
   For the errors that it can't *correct*, it'll still generate warnings
   to your system software to let you know (if you've configured it).
   
   If you do get such a warning - or if the system starts acting funny like
   you saw - that's when you'd want to run memtest on the system.
   
   ---
   
   The other time to run memtest on the system is when you first buy or
   receive a new server.  You'd generally do a "burn in" test of all the
   things (memory, hard disks/ssds, cpu, gpu, etc) just to make sure
   everything is ok before you start using it for important stuff.
   
   Regards and best wishes,
   
   Justin Clift