Details
-
New Feature
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
-
None
-
Point
-
Not applicable
Description
From Eduardo:
Per yesterday's conversation it seems like Richard would prefer that we just use the Helm chart directly rather than generating our own manifest, if that is the case, we will need at least one update. Since our goal is to run thor and roxie on different machine types, then we'll need those deployment/pod specifications to have a nodeSelector on them to target separate nodepools. Since a single nodepool can only have one machine type, we'd have a nodepool for Roxie using Standard_E8d_v4 and one for thor using Standard_L8s_v2. Then the nodeselector will make sure we schedule to the right pool.