[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[monit] Question about cpu monitoring and "expect" behaviour

From: Sébastien Wacquiez
Subject: [monit] Question about cpu monitoring and "expect" behaviour
Date: Wed, 11 Mar 2009 13:50:18 +0100
User-agent: Mozilla-Thunderbird (X11/20080724)


I have some issue to check the avaibility of my service with monit.

I want to monitor a streaming solution. My problem is that sometime, the
stream stall without being detected by the server ; the url always
return a 200 http code, but it just send his header, the "start code"
for the codec and then stall.

I can detect this easily by requiring the http response to be superior
to a pre-determined size, but there is no way to do this directly (ie,
by a specific syntax) in monit for now.

So I tryed to use something like this :
if failed host localhost port 80
     send "GET /blablabla HTTP/1.0\r\n"
    expect "(.|\n){5000}"

But it's not working. It's not working either with the "content"
checking of an http check.  After some debugging, I found that monit
seems to stop himself at the first "\0" it found ... Is this the
expected beaviour ?

My second issue is about CPU checking :
 - monit does not support the hotplug of cpus.
 - I think monit lack of a cpu unit independant of the cpu count of the
host. Lots of processes are monothreaded and we want to check when they
loop at 100% of *one* CPU. For now, if you have 4 cpus, you have to
check if your process is not at 25% of cpu, and if you upgrade or
migrate your server you'll have to rewrite all your cpu check which is
not very convenient.

Last question, about monit communication with mmonit : there is lots of
plugins for other monitoring apps that can parse the monit webpage to
centralise status (like nagios) ; I suppose that monit and mmonit share
an API that could simplify those plugins by removing the need of parsing
html data. Is this API will be stable and documented, or do you plan to
reserve it for m/monit interaction ?

Thanks !

Wacquiez Sébastien

reply via email to

[Prev in Thread] Current Thread [Next in Thread]