On Wednesday 25 July 2007, matilda matilda wrote: > >>> Max Hofer <[EMAIL PROTECTED]> 25.07.2007 12:12 >>> > > > > This is a simple assymetric co-location rule (since 2.0.8 all co-location > > rules are assymetric). > > > > This means write a rule that B follows A: > > <rsc_colocation id="B_follows_A" from="B" to="A" score="INFINITY"/> > > Hi Max, > > that is exactly what I thought, too. > BUT: With this constraint, if A is NOT started, B wouldn't be started > either. And I want B to be started in this case totally independent of A. It really would help if you just attached the CIB before and after you started resource A. ;-)
What I do when something is not woirking out as i supposed is to create a simple CIB with dummy resources which just implements the behaviour I'm intersted in (i have quitea collection of how-to-cib-examples.xml impelemnting different features). So if you have a question you can send the CIB and we can try it out and give you the answer. Back to your question. The behaviour of resource B really depends on following factors: - resource_stickiness (resource parameter) - resource_failure_stickiness (resource parameter) - fail-count of B on the different nodes - constraints which you put on B If you did not define the stickiness parameters for B keep in mind that there a default cluster values like 'default_resource_stickiness' and 'default_resource_failure_stickiness' which are default values. So it really boils down to following questions: - what is the resource score of B on all nodes when A is running - what is the resource score of B on all node when A is not running To see the resource score use ptest. Here some hints: * ptest -L -VVVVVVVV 2>&1|grep node_score will give you the score on the current life system - adapt the grep to filter out unwanted info * you can also use ptest on a stored CIB files by the heartbeat framework Check the ha-log when your resource B is stopped . You will see a log message like "PEngine Input stored in: /var/lib/heartbeat/pengine/<filename>.bz2" Use the previous stored file (where B was running) and the one shown in the log message (where B is stopped) - and check the scores and the rules affecting it before and after B was stopped * to see how the score of B is calculated use following greps ptest -L -VVVVVVVV 2>&1|grep test_expression|less ptest -L -VVVVVVVV 2>&1|grep native_rsc_location|less Once again you can do that on stored CIB files. Dig into the output of ptest - maybe you find helpfull info. > > INFINITY is only a suggection. Just make sure the co-location rule has > > high enough score so that it over powers the resource-stickiness of B (if > > you defined one - this should be the case if the resource stickiness of B > > is 0). > > Does this mean, that the docu is outdated at this point? Docu says, that at > the moment (I don't know which moment) only INFINITY and -INFINITY are > allowed. Yep outdated. You can really use any resource score. > Where do I get informations about the current (2.1.0) rules possible? As everywhere in technical life the doku is lacking behind the implementation. I think you have to wait until someone updates the Info on the page/wiki (btw - you can do that on the wiki too) kind regards Max _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
