Currently, the plugin runs a Task process which collects the metrics and publishes them to AWS CW.
This process is unique per node which means either we loose metrics publishing if the node running the process goes down or we run one process per node and we publish metrics multiple times.
We need a singleton mechanism ensuring the Task process is always run once across multi-node clusters.
noxdafox
commented
5 years ago
Currently, the plugin runs a Task process which collects the metrics and publishes them to AWS CW.
This process is unique per node which means either we loose metrics publishing if the node running the process goes down or we run one process per node and we publish metrics multiple times.
We need a singleton mechanism ensuring the Task process is always run once across multi-node clusters.
The :global or the signleton modules could be of help.