> Date: Wed, 6 Aug 2008 13:03:10 +0200 > From: "Andrew Beekhof" <[EMAIL PROTECTED]> > > As far as the cluster is concerned, the desire for a resource to stay > where it is is outweighed by the negative colocation constraint saying > that it cant. > The order in which resources are allocated depends on a) the order > they occur in the configuration, b) their priority and c) how the > colocation constraints are set up. > > Eg. > <rsc_colocation to="Gha2" score="-INFINITY" from="Gha1" > id="CLGha1Gha2"/> > Implies that Gha2 must be allocated first (so that we know where > _not_ to put Gha1) >
So if I understand correctly, there is no way using rsc_colocation and being perfectly symmetric in a symmetric cluster? I can't use colocation constraint if I want to make the cluster fully symmetric with minimum number of resources migrations on fail-over? How can I make n groups never colocate with each other and still keep minimum number of migrations on fail-over? Thanks > On Mon, Aug 4, 2008 at 21:39, Itay Donenhirsch <[EMAIL PROTECTED]> wrote: >>> Date: Mon, 4 Aug 2008 14:10:23 +0200 >>> From: "Andrew Beekhof" <[EMAIL PROTECTED]> >>> Subject: Re: [Linux-HA] nodes failover order >>> >>> can you repost your current configuration? >> >> Sure thing. See attached file generated by hb_report. >> >>> >>> On Mon, Jul 28, 2008 at 14:01, Itay Donenhirsch <[EMAIL PROTECTED]> wrote: >>>>> From: "Andrew Beekhof" <[EMAIL PROTECTED]> >>>> >>>>> On Mon, Jul 28, 2008 at 10:43, Itay Donenhirsch <[EMAIL PROTECTED]> wrote: >>>>>> First of all, thanks for the speedy reply. >>>>>> >>>>>>> From: "Andrew Beekhof" <[EMAIL PROTECTED]> >>>>>>> >>>>>>> On Sun, Jul 27, 2008 at 15:21, Itay Donenhirsch <[EMAIL PROTECTED]> >>>>>>> wrote: >>>>>>> > Hi all, >>>>>>> > I have a weird problem: >>>>>>> > >>>>>>> > There are 4 nodes: ha1 ha2 ha3 ha4 >>>>>>> > There are 3 resource groups: Gha1 Gha2 Gha3 >>>>>>> > The cluster is symmetric. >>>>>>> > Upon startup - Gha1 lives on ha2, Gha2 lives on ha3 and Gha3 lives on >>>>>>> > ha4. >>>>>>> > >>>>>>> > I do a 'service heartbeat stop' on ha3. I expected to see Gha2 >>>>>>> > migrate from >>>>>>> > ha3 to ha1 but instead I get that Gha2 migrates to ha2 and Gha1 >>>>>>> > migrates to >>>>>>> > ha1. Why is that? >>>>>>> >>>>>>> Because you didnt tell it you cared. >>>>>> >>>>>> I don't think you understood me correctly, maybe I didn't make myself >>>>>> clear enough, sorry. >>>>>> I don't want the resources to fail in a specific order, just in such a >>>>>> way that will not make non-failed resources switch nodes. >>>>>> I don't care which node gets the failed resource as long as it will be >>>>>> done in the minimal number of fail-overs. >>>>>> It seems to me that it should have worked using symmetric cluster and >>>>>> no location constraints at all. >>>>> >>>>> default-resource-stickiness >>>>> >>>> >>>> I'm sorry but I don't get it. My default-resource-stickiness is >>>> INFINITY. If I understand correctly it says how much the resource >>>> "wants" to stay where it is upon fail-back. Here my problem is that a >>>> few resources change nodes. What am I missing? >>>> >>>> In the document you refereed me to there is exactly one line about >>>> default-resource-stickiness: "How much do resources prefer to stay >>>> where they are? Used when... ". In my case "where they are" is not >>>> relevant because they all move to places they never been to... >>>> >>>> But don't I seem to understand? >>>> >>>> Thanks >>>> Itay _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
