|
From: | Jan-Henrik Haukeland |
Subject: | Re: Not the way I would hav dependents work |
Date: | Wed, 16 Feb 2005 13:05:34 +0100 |
On Feb 16, 2005, at 0:48, Chris Lopeman wrote:
SO I recently discovered that during a restart a dependent process willrestart at the same time as the parent process. Basically speaking theparent start and in less than a second (I believe before the parent script even completes), the dependent's start script is run. If these are being forked off it is even possible for the dependent script to actually complete execution before the Parent has really started execution. Anyway, seems to me that dependent should mean dependent. Monit should check or start dependent processes unless what they depend on is all good (running and responding). Instead Monit attempts to run and monitor these processes, basically setting itself for failure and errors.
You have a point here. Upon restart of a depend tree monit will do as you expect on shutdown, but maybe not on start up. Or rather the startup is debatable.
Assume this depend tree: A->BNow on restart of A monit will first stop B and *wait* on B to stop *before* it will stop A. However on startup monit will simply fork of and start B and then A, *in sequence* but without waiting for B to actually start before A is started. In other words you _can_ get the behavior you describe above. Maybe we should change the startup sequence in monit to behave as on stop? I.e. that monit waits on B to start before it attempts to start A.
-- Jan-Henrik Haukeland mobile +47 97141255
[Prev in Thread] | Current Thread | [Next in Thread] |