Search code examples
gridgaingrid-computingignite

Multiple Masters in GridGain

A few years ago, this gridgain question was posed and addressed at Gridgain failover of master (sender) node

Fast forward to today. We run gridgain 6.2.1 (open source edition) and are trying to enforce high availability of our gridgain master (in case the hardware or the master process fails). From a high level perspective, we submit (distributable) jobs to the master node via a continuous mapper manager that we define.

As per the previous accepted answer to the question, this was not possible in 2011 .

  1. Has the status quo changed in 2015, or are there other strategies available in newer versions of gridgain?
  2. Can this functionality be implemented in ignite via "cluster groups" ? See http://apacheignite.gridgain.org/docs/cluster-groups

Thanks

Solution

  • All suggestions listed in the reply are still in play.

    In addition there is] ComputeJobMasterLeaveAware interface which can be implemented by any job. Its onMasterNodeLeft method will be called on a running job if the master node leaves topology. This will allow you to save any intermediate data and resubmit jobs if needed.

    Will this work for you?