stop method: SIGKILL option

Issue #554 new
Tildeslash repo owner created an issue

In some cases it's important to kill the process on stop if graceful stop failed (such as active-passive cluster).

The question is, if if kill on stop failure should be default - it may be unwanted in some other application cases: for example database stopping slowly - hard kill may corrupt the data and the admin may want to keep the process running and investigate the stop problem rather then automatically killing the process.

It could be thus more safe to add a new option for the stop method, which will kill the process on stop failure/timeout, for example:

stop method = "/usr/bin/foobar stop" with kill on failure

To make the configuration simpler, it could be also useful to have some global options (similar to "set limits"), which will allow to enable kill for all stop methods if the user wants to, for example:

set defaults {
    stop failure: <alert | kill>   #note: "alert" is implicit, can be overridden with "kill"

Comments (0)

  1. Log in to comment