[Linux-HA] RA termination and timeout

[Andrew Beekhof]

On 2/28/07, Yan Fitterer <yfitterer at novell.com> wrote:
> >> or is it
> >> better to implement a "sub- timeout" in the RA itself, so that the RA will
> > self- terminate if an action takes longer than
> >> expected (and return the relevant error exit code) ?
> >
> > short answer: no
>
> Interested in the long answer, unless it's simply a case of:
>  - messy
>  - pointlessly redundant

pretty much.

each action already has a timeout which is enforced by the lrm.
if the internal timeout is shorter... you gave up before you needed to
if the internal timeout is longer... the lrm will shoot you before you
get a chance to exit

though lmb does have a good point about escalating shutdowns from
"nice" to "forced".
In such cases, I'd probably check the OCF_RESKEY_CRM_meta_timeout
variable (which is how long the lrm is going to give you) and use some
percentage of that.

>
> One of the reasons I started going this was to allow the RA to log its own message on timeout. I have since thought of
> using "trap" for that.  So I was then left with the question "why bother?", to which I found no good answer. I still
> dislike the look of an infinite loop in a piece of code though - probably some primitive logic-related gene  in my DNA
> ;)

its more than a little counter intuitive :-)

>
> I'm sure your long answer will be better than mine :)
> _______________________________________________
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>