On 19 Feb 2014, at 10:48 am, Kristoffer Grönlund <[email protected]> wrote:
> Hi everyone, > > On Mon, 17 Feb 2014 10:54:29 +0900 (JST) > [email protected] wrote: > >> Hi Andrew, >> >> I found your correction. >> >> https://github.com/beekhof/pacemaker/commit/37ff51a0edba208e6240e812936717fffc941a41 >> >> Many Thanks! >> Hideo Yamauchi. >> >> --- On Wed, 2014/2/12, [email protected] >> <[email protected]> wrote: >> >>> Hi All, >>> >>> There is difference in two between handling of "sequential" of >>> "resouce_set" of colocation. >>> > > I suspect that sequential for resource sets is still broken. I have > been trying to figure out what the expected value of sequential is if > not explicitly set on a resource_set. It seems like the answer is "it > depends". > > In pengine/constraints.c, unpack_order_set, sequential defaults to true > if not set, via the following logic: > > /* line 1368 */ > const char *sequential_s = crm_element_value(set, "sequential"); > /* line 1384 */ > if (sequential_s == NULL) { > sequential_s = "1"; > } > sequential = crm_is_true(sequential_s); > > However, in pengine/constraints.c(450), template_to_set, sequential > defaults to false: > > /* line 495 */ > /* Set sequential="false" for the resource_set */ > crm_xml_add(*rsc_set, "sequential", XML_BOOLEAN_FALSE); > > Am I reading this wrong, or is this on purpose? It appears Yan did this on purpose. The reason would likely be that this set is for use in a location constraint (not ordering or colocation). And in particular, it is creating a fake set for resources using the same template - so there is no reason to think they should be ordered. So: 1. sequential should _always_ default to true 2. please report it to me if you find somewhere that does not 3. since template_to_set() is creating (copying) a fake set, there is no conflict with 1.
signature.asc
Description: Message signed with OpenPGP using GPGMail
_______________________________________________ Pacemaker mailing list: [email protected] http://oss.clusterlabs.org/mailman/listinfo/pacemaker Project Home: http://www.clusterlabs.org Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf Bugs: http://bugs.clusterlabs.org
