On a single node setup, rook will fail to get stuck waiting on OSDs as
monitors by default have an anti-affinity rule that prevents multiple
monitors to land on a single node. This will also be an issue if you are
using this setup of < 4 nodes (1 master and 3 workers) as the master has
the NoSchedule taint. This patch allows multiple monitors on a single
node. When multiple nodes are present, the monitor pods are
automatically distributed.
Downside: When the setup is started with a single node and then more
nodes are added, the monitor pods DO NOT automatically re-balance.
On a single node setup, rook will fail to get stuck waiting on OSDs as monitors by default have an anti-affinity rule that prevents multiple monitors to land on a single node. This will also be an issue if you are using this setup of < 4 nodes (1 master and 3 workers) as the master has the NoSchedule taint. This patch allows multiple monitors on a single node. When multiple nodes are present, the monitor pods are automatically distributed.
Downside: When the setup is started with a single node and then more nodes are added, the monitor pods DO NOT automatically re-balance.
Fixes: #306
Signed-off-by: Ganesh Maharaj Mahalingam ganesh.mahalingam@intel.com