I use webminstats as a diagnostic tool, to understand why a host is bad/not working. Here are some real cases, I encountered :
file system full
I use a package called
diskcheck to have alarms on almost full file systems (very helpful tool).
But webminstats (disk module) can show if it's a slow growing (bad cleaning ?) or a file explosion (core ?).
The filesopen module can then help to know if there was an abnormal number of open files.
not enough memory
A server was responding very very slowly (almost impossible to connect it). On console, syslog showed "not enough memory". So I used the magic keys to sync my disks and reboot the server.
Then, I use webminstats to investigate. The memory module shows a quick grow of swap and used memory.
A look at process module shows another big grow, and the internet module (number of process) shows the culprit : the number of java process growed from 100 to 1400.
high cpu load
On a bi-proc server, a regular check on the cpu module shows me 100% cpu (max 200) used for a week.
I can see only one processor was used (line graph mode). Nobody notice the problem, because the second processor was free.
I just had to use top to find the process, and killed it.
apache not working
A user inform me, that a web server was not working. I had a look at internet module,
and I remark the httpd activity (seen throw apache syslog files) just fall down.
The user module shows a connection just before the problem, the who command identifies the person,
and ls command shows the apache config file was just modified ...
inetd configuration
On a pop server, I had problems with inetd default configuration, which stopped the service if two many connections
(protection against dos attacks). So I put a very high filter, then reduce it with statistics from the mailq module.
where is the bottleneck ?
A user complains about bad response time on a server. A turn on webminstats (cpu, network, memory, irq for disk I/O) may help to identify the problem.
back to main page
Page changed on 6 March 2003