Merge pull request #215 from grafana/parallelly-scale-compactor

Switch compactor statefulset to Parallel pod management policy

Author: pracucci

CHANGELOG.md

@@ -14,6 +14,7 @@
   ingester_statefulset+:
       statefulSet.mixin.spec.withReplicas(6),
   ```
+* [CHANGE] The compactor statefulset is now configured with the `Parallel` pod management policy, in order to scale up quickly. #214
 * [ENHANCEMENT] Add the Ruler to the read resources dashboard #205
 * [ENHANCEMENT] Read dashboards now use `cortex_querier_request_duration_seconds` metrics to allow for accurate dashboards when deploying Cortex as a single-binary. #199
 * [ENHANCEMENT] Improved Ruler dashboard. Includes information about notifications, reads/writes, and per user per rule group evaluation. #197, #205

cortex/tsdb.libsonnet

@@ -163,7 +163,12 @@
     statefulSet.mixin.spec.selector.withMatchLabels({ name: 'compactor' }) +
     statefulSet.mixin.spec.template.spec.securityContext.withRunAsUser(0) +
     statefulSet.mixin.spec.updateStrategy.withType('RollingUpdate') +
-    statefulSet.mixin.spec.template.spec.withTerminationGracePeriodSeconds(900),
+    statefulSet.mixin.spec.template.spec.withTerminationGracePeriodSeconds(900) +
+    // Parallelly scale up/down compactor instances instead of starting them
+    // one by one. This does NOT affect rolling updates: they will continue to be
+    // rolled out one by one (the next pod will be rolled out once the previous is
+    // ready).
+    statefulSet.mixin.spec.withPodManagementPolicy('Parallel'),
 
   // The store-gateway runs a statefulset.
   local store_gateway_data_pvc =