On Mon, Mar 18, 2019 at 07:38:48PM -0700, Han Zhou wrote: > From: Han Zhou <[email protected]> > > Add option --leader-timeout to ovsdb-server. > > Signed-off-by: Han Zhou <[email protected]> > --- > > Notes: > This patch is on top of: > https://mail.openvswitch.org/pipermail/ovs-dev/2019-March/357339.html
It seems like a reasonable idea to make this configurable. I am not sure that this is the right way to configure it. It might be better as part of the Raft storage data structures, the ones that are stored on disk with database itself. This could be especially important if a given ovsdb-server serves more than one clustered database, if it's possible that different databases have different settings. I think that \fIserver\fR should be \fItimeout\fR or similar here, and I guess that "miniseconds" should be "milliseconds": > +\fB\-\-leader\-timeout=\fIserver\fR > +Set the cluster leader election timeout value in miniseconds. This option > +should be set to the same value across all nodes in the cluster. Leader > +re-election will be triggered if a follower haven't heard from current leader > +within this interval. The default timeout is one second. Increasing this > +value reduces the chance of leader re-election during transient overload > +situations but increases the delay of reacting to real failures, too. Thanks, Ben. _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
