Hello everyone! I have great news for all of you! It seems your support has found what was causing the CSP locks. There were actually two different problems that caused the lock: 1) an upgrade of Cach� while CSP still remained version 4.1 2) a wrongly advised Kernel tuning
I'll explain these two issues in detail bellow: 1) CSP version 4.x on Cach� 5.x When we upgraded our development box to Cach� 5, the CSP package were not upgraded, so we were using a version 4 CSP daemon and CGIs with a version 5 superserver, which caused eventual locks. Enough said about this one. 2) Wrongly advised kernel tuning Our production box was formatted and had Cach� and RedHat installed from scratch, so the above could never cause the locks. Instead, the Cach� installer advises you to increase a shared memory parameter (SHMMAX) which is said to be too low for Cach� applications when set to the standard value of 32MB on Linux boxes. HOWEVER, all other documentation about kernel tuning tells of another parameter (SHMALL), which USUALLY appears with the same value as SHMMAX. Problem is, the installer simply DID NOT mention this parameter was to be changed!!!!! So, if you follow Cach�'s advice and set SHMMAX to 256MB for example, you would have SHMMAX=256MB and SHMALL=32MB. We changed SHMALL to 256MB on Friday and Cach� has been running smoothly ever since :) Anyway, the above mentioned tuning is in the Cachpe docs as well as RedHat's kernel tuning page (actually, the former is copy-paste of the latter....). Here is the RedHat link http://www.redhat.com/docs/manuals/database/RHDB-7.1-Manual/admin_user/kerne I hope this helps you guys :) []s F�bio "F�bio Gusm�o Ribeiro" <[EMAIL PROTECTED]> escreveu na mensagem news:[EMAIL PROTECTED] > In our case memory seems alright, and the webserver outputs HTML documents > just nicely, as if nothing ever happened........ But if you look closer, > there're 5 or more CSPnsd processes running at the same time and no response > when a csp is requested! > > Also, I heard the Linux cron hangs as well......could the two be related? > > > > > > "Nuno Canas" <[EMAIL PROTECTED]> wrote in message > news:[EMAIL PROTECTED] > > I had similar behaviours, even cache consuming 98% of cpu because of > > some "bug" or bad coded csp. > > There are some processes known in unix as "runaways" that never stop and > > keep consuming resources forever. > > Depending in what size of shared memory segment you have it's easy to > > get out of swap space (another good reason to stop reponding). > > If it's CSPnsd YES, it's CSPgateway, if not, if it's cache process it > > can be something derived from errors in code that generate a big > > pipelining and freeze the machine. > > Another interesting point is to check what's frozen: process (user), or > > machine (system) and see if you kill the process the machine recovers or > > not. > > > > Nuno > > > > Denver Braughler wrote: > > > > > Nuno Canas wrote: > > > > > >>You're saying it's CSPGateway, so is it CSPnsd daemon that is frozen !? > > > > > > > > > When there are crashes, odd and apparently unrelated behaviors can > arise. > > > I had trouble with Studio and Terminal pursuant to a crash during a CSP. > > > Terminal would appear fine, but freeze up on a HALT. > > > >
