[Linux-HA] failcount isn't increased when a resource fail to start

[Andrew Beekhof]

On 7/30/07, Junko IKEDA <ikedaj at intellilink.co.jp> wrote:
> Hi,
>
> This is a topic that was talked about here before,
> could I confirm it again?
> my concern is;
> failcount isn't increased when a resource fails to start.

this is something that will be changed "soon"
start failures are (today) treated as a special case in that the
resource simply cant run on that node anymore... so updating the
failcount seemed somewhat redundant

> a failure status of start op can be cleaned using 'crm_resource -C',
> but using both 'crm_resource' and 'crm_failcount' depending on the status
> leads to a slight confusion.
> I think administrators should master it,
> it's just confined to users or operators.
> I don't understand what will be going on if start-fail increases a
> failcount,
> could you please explain it again?

the idea is that eventually we'll remove the special start-failure
logic in the PE and instead set failcount to INFINITY.  which achieves
the same result in the end, but is a little cleaner to implement and
cleanup (ie. with crm_failcount).

> anyway, I took notice of the new message means 'start operation failed' on
> crm_mon in the latest version.
> not bad at all!

glad you like :-)