On Mon, Jun 15, 2009 at 11:02 AM, Marcin Krol<[email protected]> wrote: > Serge Dubrouski wrote: >>> - what you mean by "clones" - you mean identically configured nodes? >>> That's not a problem, really, >> >> I mean the CRM feature that is called "clones". See CRM documentation. > > Sure, except where can I find it? Is there any? I googled and googled > and nothing except Pacemaker's CLI manual came up ( > http://www.clusterlabs.org/mediawiki/images/8/8d/Crm_cli.pdf )
This one is much better: http://www.clusterlabs.org/mediawiki/images/f/fb/Configuration_Explained.pdf > > Sure that's interface and that's smth, but I really have problem with > finding some sort of tutorial or book that would explain concepts and > advanced configuration for CRM and other key pieces of Linux-HA? > > I ordered Karl Kopper's "Linux Enterprise Cluster" book -- but if > there's anything, online or printed, that would explain *in detail* > advanced configurations like those, please let me know (anybody out there). I doubt that that book has any information on Heartbeat/CRM/Pacemaker. At least the edition that I have doesn't. > >>> - resource supporting active-active configuration; do you mean like >>> Apache instances in 2 resource groups configured to use different IP >>> addresses so they would not conflict? >> >> I mean that Apache should have a shared filesystem underneath it or >> some kind of another mechanism that will make sure that both instances >> have the same content. Also if you try to make an active-active >> filesystem you have to make sure that it support concurrent access on >> different nodes, i.e. you have to use something like OCFS2 or so. > > Hmm I was thinking rather of 2 different iSCSI targets or NDB devices, 1 > for each resource group -- is switching them possible on failover / > failback? Then it's not active/active is it? > > Cluster filesystems seem big, complex and have problems with performance: > > http://hoston.it/blog/2009/03/19/htaccess-revisited/ > > Not to mention things like metadata performance (just 1 Lustre metadata > server for entire fs) and if Lustre metadata server goes on vacation, I > can kiss entire cluster goodbye.. > > Anyway, is OCFS2 an answer safety-wise and performance-wise? Suppose I > wanted to run MySQL instances on this? OCFS2 can be used for shared filesystem and as far as I know Oracle recommends it for using even in Oracle RAC, MySQL is a different thing though. I don't know how to build an active/active MySQL cluster and I'm pretty sure that it requires more that a shared FS, the database itself should support concurrent access from different nodes to its data. > > Sorry for all these questions but I really need info on such vital > issues and google is not very helpful this time around.. > > Regards, > mk > > > > >> >>> Regards, >>> mk >>> >>> >>>> On Mon, Jun 15, 2009 at 10:16 AM, Marcin Krol<[email protected]> wrote: >>>>> Hello, >>>>> >>>>> Let the newbie ask another question: >>>>> >>>>> Can both nodes in active-active configuration run the same *type* of >>>>> resources? I mean, suppose I configure two separate instances of Apache >>>>> and relevant resource groups for them, can I run them on both nodes at >>>>> the same time, have one node fail over to 2nd node and have both >>>>> resource groups functioning properly on that 2nd node? >>>>> >>>>> Regards, >>>>> mk >>>>> _______________________________________________ >>>>> Linux-HA mailing list >>>>> [email protected] >>>>> http://lists.linux-ha.org/mailman/listinfo/linux-ha >>>>> See also: http://linux-ha.org/ReportingProblems >>>>> >>>> >>>> >>> _______________________________________________ >>> Linux-HA mailing list >>> [email protected] >>> http://lists.linux-ha.org/mailman/listinfo/linux-ha >>> See also: http://linux-ha.org/ReportingProblems >>> >> >> >> > > _______________________________________________ > Linux-HA mailing list > [email protected] > http://lists.linux-ha.org/mailman/listinfo/linux-ha > See also: http://linux-ha.org/ReportingProblems > -- Serge Dubrouski. _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
