Re: [Wrapper-user] Filters: triggering on things other than strings

[Leif M. <le...@ta...>]

Aiman,
See below.

Wrapper User wrote:

>As I understand it now, the wrapper can only trigger on an output string.
>
>Is there any way (or near-future plans) to trigger on something other than
>output. For example, if the program throws an Exception, or trigger on the
>exit code of the application?
>
>In my specific case, I'm having an application "service1" (which I'm under
>the assumption can't be changed to display different output) tries to
>connect to a (naming) server "service0". Note that service0 will also be
>launched by a wrapper. So, basically, service0 should be started before
>service1. But I'm starting them with the init.d services in Linux, so that
>can't be guaranteed. The way service1 works is that, upon starting up, it
>will try to connect to service0. If service0 is not running, service1 will
>retry to connect a number of times before giving up and exiting with exit
>code 1.
>
>When it can't connect to service0, service1 prints the following:
>----------------------------------------------------------
>jvm 1    | [ Retrying to connect to 127.0.0.1:1234 ]
>...number of times...
>jvm 1    | [ Retrying to connect to 127.0.0.1:1234 ]
>jvm 1    | [ Retrying to connect to 127.0.0.1:1234 ]
>wrapper  | <-- Wrapper Stopped
>----------------------------------------------------------
>
>So as you see, there's really no unique string it prints out that I can
>trigger on after it failed n times. I found it exits with exit code 1 by
>running the wrapper with logging set to debug.
>Anyway, in theory, at the same time, a different wrapper instance could (in
>my case should) be off starting service0, so I'd like service1's wrapper to
>restart service1 with the hope that service0 will eventually be up to
>connect to. Allowing the wrapper to trigger on service1's exit code would
>help.
>As a workaround (kluge?) , I could just trigger on the "Retrying to connect"
>string, but that would restart after each failed connect, as opposed to
>after the program exits in failure.
>  
>
Good news.  The ability to trigger on a JVM exit code was added in 3.1.0.
Take a look at the docs and let me know if this will do what you need.
http://wrapper.tanukisoftware.org/doc/english/prop-on-exit-n.html

>Another (totally separate) solution would be (and I'm just brainstorming
>here) putting some dependencies on different wrapper applications. Something
>like writing a totally separate script that somehow checks if service0 is
>running (service0 status) before it launches service1...
>  
>
This is available on Windows.   But UNIX does not really provide a 
mechanism to do
this other than specifying startup and shutdown orders in the various 
run levels.

I am always open to great ideas for how to do things.

Cheers,
Leif