Open jdef opened 9 years ago
@kozyraki is this what you were talking about the other day?
@jdef More or less. I can see a number of ways to do this. My preferred way would be that the points to one or more ZK instances that are initially active. Then the detector or mesos-go library keeps track of how other instances come and go into Zookeeper dynamically and accesses them for info like the leading master. That way, ZK can evolve (get updated, maintained etc) without the client ever knowing anything.
@jdef: Is this covered by #162? Also, why is all of this ZK stuff in this repository? Isn't it completely orthogonal to Mesos?
ZK detection is used to detect a change of master leadership. needed so that the scheduler driver can re-register with a new master. the ZK detection stuff could probably be a standalone project, outside of mesos-go, since it has usefulness on its own for projects that have a similar need to detect master failover.
On Thu, Sep 10, 2015 at 8:53 AM, Tomás Senart notifications@github.com wrote:
@jdef https://github.com/jdef: Is this covered by #162 https://github.com/mesosphere/mesos-go/pull/162? Also, why is all of this ZK stuff in this repository? Isn't it completely orthogonal to Mesos?
— Reply to this email directly or view it on GitHub https://github.com/mesosphere/mesos-go/issues/108#issuecomment-139227016 .
the ZK detection stuff could probably be a standalone project, outside of mesos-go, since it has usefulness on its own for projects that have a similar need to detect master failover.
My thoughts exactly. I'll create an issue for that.
Allows clients to update their server list automatically, without a complete client restart: