We set up the docker swarm cluster with dedicated CPU 4Gb hosts. In operations the available memory proves so far to be a bigger limitation than the cpu performance: when services need to relocate to a different node, having not enough memory can cause services getting stuck on node and/or become unavailable.
@fredvd has added a fourth worker with 8Gb ram shared CPU. We could add 2 more of these hosts and remove the 3 decidated worker nodes, or try to convert the existing dedicated CPU nodes to shared ones with double the RAM.
We set up the docker swarm cluster with dedicated CPU 4Gb hosts. In operations the available memory proves so far to be a bigger limitation than the cpu performance: when services need to relocate to a different node, having not enough memory can cause services getting stuck on node and/or become unavailable.
@fredvd has added a fourth worker with 8Gb ram shared CPU. We could add 2 more of these hosts and remove the 3 decidated worker nodes, or try to convert the existing dedicated CPU nodes to shared ones with double the RAM.