Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
On Mon, May 24, 2004 at 10:41:52AM +0200, Helmut Wollmersdorfer wrote: > Lars Marowsky-Bree schrieb: > >On 2004-05-23T23:26:50, > > Helmut Wollmersdorfer <helmut.wollmersdorfer at gmx.at> said: > > >>Would it be a solution, to print a message and let states as they are? > >>Further steps should be done by the cluster manager. > > >The problem is that the existing cluster managers cannot handle this > >scenario. > > Ok, misunderstanding. I used the term "cluster manager" including both, > cluster management software AND human cluster administrator. > > Back to discussion: > I agree, that there always can be a situation, which needs human > intervention, or the admin wants to configure it this way. > > What is happening now: > The init-script waits, until the admin gives an answer. In the meantime > nothing happens, states are left as they are. > > The critics are not, _that_ this is necessary. The critics are on the > form, _how_ this is done. The _init-script_ should not wait. you have the choice. if you configure it, it will do things differently. give it inittimeout=-90 and blame heartbeat if it decides to make the outdated node primary. Lars Ellenberg