[LON-CAPA-admin] Serious troubles with lonmaxima

[Guy Albertelli]

Hi Stefan,

> Back from some days of holiday I experience still the same issues with
> server overload caused by maxima. We use LON-CAPA 2.4.1. Since some
> weeks now we have a few authors creating and testing LON-CAPA problems
> which use <mathresponse>, <formularesponse> without sampling points
> and/or the new perl/LON-CAPA function &cas.
> 
> What is going wrong?
> #1: problems using &cas: It happens appr. every 20-30 new problem
> variations, that the server stops working and cancels the process after
> the default 10 seconds ("code run too long" message). Re-click on "Reset
> Submissions" works fine, no stop. Hence it's not a problem with a bad
> choice of randomized values.

I'd guess that this due to the maxima process maintenace script
become decoupled from the underlying maxima process.

My only real guess at fixing this is likely by uncommenting the line

   #$command->log_file("$execdir/logs/lonmaxima.session.log");


And when seeing this. saving away the log file and steping through to
figure out what has gone awry.


> #2: From time to time lisp processes overload the server. They use each
> more than 99% of CPU (2 x 3 GHz and nearly no users working). "Server
> Load" shows 150%, 550% or even 1700% (!!!)

Hrrm, I'd need to get access to one of the machines that is doing
this. We haven't been seeing this here at all after the changes I put
into 2.4.1 (which was to kill off the maxima sub-process whenever we
try to cleanup/restart etc. the process.)

My guess is that it's still something releated to this, and the
reaping needs to be improved in some way, but I'd need to see a
machine failing inorder to figure out what needs improving.

> #1 and #2 occur on SLES9 machine, #2 also occurs on the FC5 machine. I
> am not sure about a connection between #1 and #2. More tests will follow.
> 
> Any ideas what could go wrong?
> Are there any changes with maxima in LON-CAPA 2.4.2? Is it worth
> updating (concerning this issue)?

No no changes in 2.4.2 with reagrds to this (there is the fix for
maxima evaluation of formulareponse)

-- 
guy at albertelli.com   0-7-0-9-27,137