于 2008年10月25日 02:08, Darren Reed 写道: > On 10/23/08 22:37, Huafeng Lu wrote: >> 于 2008年10月24日 06:42, Darren Reed 写道: >> >>> - the project as it is appears to deliver a single VRRP daemon that >>> serves all groups/instances that the host is participating in. Thus a >>> change in the configuration of one has the potential to affect all >>> others. Given this is a high-availability solution, is this wise? >>> >> >> How will the change of one's configuration affect others'? >> > > If a change requires vrrpd to be restarted, it has to unjoin all of the > groups/instances it is in. Or maybe there is a problem with one of > the VRRP groups/instances that requires the daemon to be restarted. > In such a case, doing it for one impacts all.
The vrrpd daemon doesn't "join" or "unjoin" groups/instances. When vrrpd is started, it creates vrrp instances/groups according to the configuration file, if it exists; later the user can use vrrpadm to ask vrrpd to create new instances/groups, destroy/modify/startup/shutdown existing instances/groups. Changes to one instance/group don't affect other instances/groups, and no changes/operations require vrrpd to start. (We're still thinking about your comments about using SMF service instances. If they're used, the above statement may change, though.) We'll add some explanation to the design doc to make it clearer. > Which highlights another point - behaviour of the service with respect > to refresh and restart needs to be discussed in the design document. Sure. Will add some discussion to the doc. Thanks. -- Huafeng _______________________________________________ networking-discuss mailing list [email protected]
