fix broker select error in arbiter

[andyxning]

This pull request will fix the problem proposed in #1562, for more info please go to #1562。

[naparuba]

Seems valid :)

I'll give a test.

Just a pep8 error (look at https://travis-ci.org/naparuba/shinken/jobs/57749243 for it).

I'm waiting the 2.4 release before merge it, but when 2.4 is release and I test your code, will be merged :)

[andyxning]

@naparuba , i will check the pop8 error and fix it soon.

[naparuba]

Cool thanks :)

On Thu, Apr 9, 2015 at 9:43 AM, Ning Xie [email protected] wrote:

@naparuba https://github.com/naparuba , i will check the pop8 error and
fix it soon.

—
Reply to this email directly or view it on GitHub
#1565 (comment).

[andyxning]

One more thing, i do not take care about whether the push_broks function will succeed. However, in the original implementation the broks should only be removed when push_broks will succeed. I have thought that there will be some problem if we continue to collect broks when all the brokers down, thus we will continue to collect broks and consume the memory. It is dangerous only if we can make a threshold that we can not store more than high_water_level broks( or anything like that) .

If we just do not care about the execution status about push_broks then we just have some status broks lost and everything about schedule, check and notification will still work properly. When the downed broker daemon coming start again we can sent it with the newest status broks.

Since we can get the satellite link info from the update broks not the initial broks, then we can simply send the satellite link broks and then remove all of them from self.broks. This will not have impact on the correct behavior both in arbiters and brokers. :)

Additionally, if the broker down unexpectedly, then we can have problems when we restart it and it can not have any information about the satellite links, so it will not display the System info in WebUI's System tab.

[andyxning]

@naparuba , can you please take a look at this PR.