[Dxspider-support] Weird node issue that prevents user logins

[Brendan Minish]

On Thu, 2009-10-08 at 22:55 +0100, Dirk Koopman G1TLH wrote:
> Brendan
> 
> What version of perl are you using and is it the same on both boxes? 
> What does the other box say was passing through at the time? You have 
> done the 'perl user_asc' thing and regenerated your user file?

It's perl V 5.8.8 on both boxes as packaged by centos

yes this problem has survived having the user database rebuilt on a
number of occasions. however this time around I have gone back to a much
older user_asc from a backup from just before the move to see if that
helps matters any

your ssh key is on the box if you want to have a poke about 

> Could it be hardware?

I guess it's possible but it's a XEN instance running on a much bigger
server owned by tagadab.com. I suspect that if there was a dogey ram
problem etc it would get noticed fairly quickly because it would be
hitting many others as well ? 

 
> Google has lately come up with a rather worrying picture of rather high 
> DRAM error rates which seem to be related to M/B design.

Yes I read that too. I also ran into a google employee once who told me
about coming across an issue with ram (that took a long time to
diagnose) where they had to send back pallets of the stuff. apparently
after the vendor tested it they then barcoded it for google but the
labelling machine was off kilter and put the bar code labels on with too
much force creating a lot of faulty ram in fairly short order. 

> It is a source of considerable concern to me that many machines are 
> totally reliable (GB7DJK springs to mind) and a few fall over rather 
> frequently. Usually corrupting the user file on the way. It appears that 
> with 5.10, things have become noticeably worse and I can't say that I 
> recommend it at the moment. But some people don't have a choice.

ei7mre never gives me any trouble and if only I could see where the
differences are.
EI7SDX has more users is it possible that we could be running out of ram
on occasion? This XEN host has 256megs but it usually seems to run with
over 100Megs free 

> DB_File together with Berkeley DB is, frankly, a pile of sh*t*, but the 
> alternatives are either just as bad (but in different ways) or require 
> extra packages which usually difficult to find for Windows users.

mysql? 

> It is starting to become a bigger worry.
> 

-- 
73
Brendan EI6IZ