A common pattern for Helix controller's high-availability is that you could set up a "controller cluster" (typically a cluster with 3 nodes or so) with the LeaderStandby state model. That way, you guarantee that every cluster gets a Leader controller with two backup Standby controllers.
Hunter On Sun, Oct 20, 2019 at 9:01 PM kishore g <[email protected]> wrote: > It happens automatically. > > On Sun, Oct 20, 2019 at 8:27 PM Jianzhou Zhao <[email protected]> wrote: > >> cool, is it done automatically by helix? or set by any user option? >> >> On Sun, Oct 20, 2019 at 8:25 PM kishore g <[email protected]> wrote: >> >>> It’s not a SPOF. You run multiple controllers and Helix ensure that >>> there is only one controller active at any given time >>> >>> On Sun, Oct 20, 2019 at 8:20 PM Jianzhou Zhao <[email protected]> wrote: >>> >>>> Hi, >>>> >>>> Is the controller a single-point failure or replicated with backup >>>> instances? >>>> >>>> Since the controller decides all transitions w/ monitoring all >>>> clusters, what happens if the controller fails? >>>> >>>> Thank you, >>>> j >>>> >>>
