Am Freitag, 22. Dezember 2006 23:04 schrieb Jim Fulton:
> An abnormal exit is an exit due to some error, such as a segmentation
> fault or running out of memory.
> > in the sense, how would the
> > supervisor know thats it's abnormal?
> It can check the exit status. Normal shutdowns typycally exit with
> status 0. zdaemon (and supervisor2, which was obviously inspired by
> zdaemon) has an exit-codes option to configure which exit codes
> are normal. If the supervisor exists with one of these exit codes,
> it should not be restarted.
i see. however, i guess the idea of a supervisor is just for such cases: to
take care to restart a server that broke. there's "svc -o" to run a service
only "once", which comes handy if one is not sure if the server will survive
a start. what i think to make more sense would be that the supervisor could
detect if a server is restarted too frequently. in my experience, thats more
often a situation that needs special attention.
> > FWIW, there's also "runit", obviously developed in the same spirit but
> > independend of daemontools: http://smarden.org/runit/useinit.html
> This wants to run as process 1. This seems a bit intrusive.
yes and no. the page i linked to explains how to use it while _not_ replacing
init. in that mode it seems to be quite like daemontools (see "The sv
program", "The runsvdir program", "The runsvchdir program" and "The runsv
program" on the main runit page). i guess there is at least some support for
your use case, see http://smarden.sunsite.dk/runit/runsv.8.html: one can
define a "finish" script, which is started with the exit code of the
supervised process as one of its parameters.
Lalisio GmbH www.lalisio.com
Puschkinstraße 1 fon +49-(0)361/541 43 80
99084 Erfurt fax +49-(0)361/541 43 79
Zope3-dev mailing list