Hi! Coming from PA-RISC and HP Integrity Servers with a Swicthover and Serviceguard background), I'd say if a server is healthy enough to boot, it should try to join the cluster: If it's able to join the cluster, it should be healthy enough to run resources. Eventually it's the task of the cluster software to find out which resources may run.
IMHO most cluster problems are software or configuration problems, not hardware problems. Booting will probably not fix any of those, but maybe it will make them obvious. Regards, Ulrich >>> <[email protected]> schrieb am 09.08.2012 um 09:27 in Nachricht <of6629794c.1418036f-onc1257a55.002709dd-c1257a55.0028f...@bull.net>: > Hi, > > I do not agree, that's not because we have to start manually HA soft on a > rebooted node that we > can say "manual HA" ... I've already had lots of configurations with HA > solution, from HACMP (IBM) , > Cluster Suite (RH) and finally Pacemaker, and none of users wants an HA > auto start at boot time > for an obvious reason : if it is a breakdown that brokes the node after > each boot (i.e a little HW problem, etc.) > it will always be fenced a short while after reboot and at worse may > disturb other node(s) which are always > working . So in this case, you can't say "better to fix the problem which > leads to repeated reboots" > as before the breakdone there was no problem ... that's why I personnally > (and it is always recommended) > never recommend to start HA at boot time. > So I'm not fan of "HA autostart " ;-) > > Alain > > > > > De : "Ulrich Windl" <[email protected]> > A : "General Linux-HA mailing list" <[email protected]> > Date : 09/08/2012 09:04 > Objet : [Linux-HA] Antw: Re: open-iscsi won't automatically log in to set > up SBD device > Envoyé par : [email protected] > > > >>>> Dejan Muhamedagic <[email protected]> schrieb am 08.08.2012 um 17:20 > in > Nachricht <20120808152017.GG3580@squib>: > > [...] > > It is arguably good practice that nodes don't join cluster > > automatically on boot. Normally, one wants to investigate why > > the node rebooted. > [...] > > Hi! > > Why shouldn't it be possible to find out why a node rebooted when it's > automatically joining the cluster? > > Unless there is a problem that repeatedly causes the node to boot, it > should be OK to automatically join the cluster. Otherwise one might get > quorum problems. > > IMHO it's better to fix the problems that cause repeated reboots, rather > than make cluster startup manual. Isn't "automatic" a prerequisite for HA? > ;-) > I'm not a fan of "manual HA". > > Regards, > Ulrich > > > _______________________________________________ > 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
