For users deploying to clusters of mixed compute types (e.g. on-demand vs spot) it is helpful to be able to force applications to skew across particular topology characteristics so not all replicas are wiped out if all spot instances are terminated.
Just like tolerations it should be possible to configure these, e.g.:
topologySpreadConstraints:
- maxSkew: 1
topologyKey: "karpenter.sh/capacity-type"
whenUnsatisfiable: ScheduleAnyway
This should then be used throughout the chart with appropriate labelSelectors inserted as required.
For users deploying to clusters of mixed compute types (e.g. on-demand vs spot) it is helpful to be able to force applications to skew across particular topology characteristics so not all replicas are wiped out if all spot instances are terminated.
Just like tolerations it should be possible to configure these, e.g.:
This should then be used throughout the chart with appropriate labelSelectors inserted as required.