[Shinken-devel] How to do a reload in Shinken

[Andy Xie]

Recently, our team want to make use Shinken to monitor our services. 

In our situation, we need to modify the configuration a little bit more. Every time the configuration change we should user the reload function provided by the Shinken init.d script which  will turn out to stop arbiter and start it again. a little bit more on reload of Shinken with this will make arbiter unavailable for some time. 

However, if we do not reload the configuration as soon as the configuration changes we can lose some thing if the service is not OK. So, what schema will mostly you chose to do a reload and what make you choose that schema. :)

++++++

Ning Xie

[nap]

Hi,

There is no perfect solution currently for this. One problem is that the current init.d script is not using the best solution for the arbiter case. When you reload, one big part of the time is used to read the configuration, parse/create/link objects, and finally serialize them (that part is maybe the biggest one). 

But all of this do not need the old arbiter to be killed (you can have the old one still running and the new working and consumming the CPU for the serialization).

We should change the reload call to not stop the old arbiter, but instead call the new arbiter with the -r option: when going to daemon (so after the load phase if I'm not wrong) that's the new arbiter that will kill the old one. So it will be killed "only when need" ^^

You can give a try by hand (-d and -r so for your new arbiter) and see the effect on the "arbiter off" duration ^^

Jean

 

++++++

Ning Xie