Petteri Räty <[email protected]> wrote:
> On 11.12.2013 15.54, Michael Fischer wrote:
> > On Tue, Dec 10, 2013 at 10:44 PM, Petteri Räty <[email protected]> 
> > wrote:
> > 
> >> At least for pid based monitoring tools it is (I do agree with others
> >> that you should also be monitoring http though). For example monit
> >> requires that you give it a pid file. Why is it wrong for them to point
> >> to the same pid?
> > 
> > Monit doesn't require a pid, never has:
> > 
> > http://mmonit.com/monit/documentation/monit.html#connection_testing
> > 
> 
> If you only want to do connection testing. What if you want to monitor
> the memory usage of unicorn processes?

Your monitoring has to adapt to the new processes anyways.  Can it
really not deal with a PID file being temporarily absent?

There's a plethora of tools in the wild which deal with PID files.
Consider this case:

It's likely a tool will see foo.pid.oldbin existing, and wait for
foo.pid to become available.  If foo.pid is already available from a
hardlink, it'll be pointing to the old PID, and any tool which backs out
of the upgrade by intending to kill the new PID will end up hitting the
old one.
_______________________________________________
Unicorn mailing list - [email protected]
http://rubyforge.org/mailman/listinfo/mongrel-unicorn
Do not quote signatures (like this one) or top post when replying

Reply via email to