[Linux-ha-dev] Re: [Linux-ha-cvs] riloe plugin commit

[Lars Ellenberg]

/ 2004-10-01 18:23:40 -0500
\ Guochun Shi:
> >
> >First, a general criticism.  (OK. I know I'm flogging a dead horse here.
> >But. I can't help it.) I thought the idea for STONITH modules would be
> >that they can be loaded as plugins proper and thus the fencing doesn't
> >need to touch the disk. I still maintain that it is either that - and
> 
> I did not know the initial design but any specific reason not to touch the disk? 
> The machine stonithing the other is the good one and its disk is expected to be good 

the reasonning goes along the lines
 stonith is vital, and all other things depend on it,
 so it may not block for any reason, so it should be locked in memory,
 and not touch code pathes that may block.
 and not touch the disk, since the IO path may depend on fencing...

the not that paranoid would say: so what, it needs to use the network
anyways (if it is a network power switch), and that may block, or ...
or if this box cannot swap a single page, I want it to drop out of the
cluster rather sooner than later anyways... 
 :-/

> >And, as some more hands-on feedback, the RILOE_COMMAND is hardcoded to
> >/usr/lib/heartbeat/, which is not where heartbeat will end up on a 64bit
> >architecture (or if the admin overrides the path). Please use the proper
> >defines.
> 
> I will fix that hard-coded path. Thanks for pointing that out.

and while at it, in linux we compile with -Werror -Wformat=2,
and that means that the warning issued for non-constant format string
arguments to printf like functions is then turned into an error and ...
(which is ok, or I'd "only" need to write some bogus translation with
some more format characters in it, and all goes boom)

what I wanted to say, maybe better do it like this:

diff -u -p -r1.1 riloe.c
--- lib/plugins/stonith/riloe.c	1 Oct 2004 20:12:49 -0000	1.1
+++ lib/plugins/stonith/riloe.c	2 Oct 2004 05:17:03 -0000
@@ -311,7 +311,7 @@ riloe_reset_req(Stonith * s, int request
 		syslog(LOG_ERR, "invalid argument to %s", __FUNCTION__);
 		return(S_OOPS);
 	}
-	syslog(LOG_INFO, _("Host %s riloe-reset."), host);
+	syslog(LOG_INFO, "%s '%s'", _("riloe-reset host"), host);
 
 	sprintf(cmd, "%s %s reset", RILOE_COMMAND, host);


	lge