Spinnaker Pipeline expression examples, tips, and tricks
Note: This guide assumes you’re deploying Spinnaker on Kubernetes using the Distributed deployment type with Halyard.
You may find the need to apply custom limits and requests to some of the Spinnaker services – doing so can provide more stable behavior of the services while also preventing extreme bursts of CPU activity or the allocation of more memory than is available (and thus being OOM Killed).
Custom sizing can be applied to a service through the
deploymentEnvironment.customSizing section of your hal config. The following example sets requests for
gate of 250 millicpu and 512 MiB with limits of 500 millicpu and 1GiB of memory.
deploymentEnvironment: customSizing: gate: limits: cpu: 500m memory: 1Gi requests: cpu: 250m memory: 512Mi
(Note: for these settings to be applied to sidecar containers as well, prefix the service name with “spin-“)
For a more complete understanding of these settings see the Kubernetes docs
In addition to the options available through the
deploymentEnvironment.customSizing settings, JVM-based services have
JAVA_OPTS that can be set or overridden through their service-settings. This is done by creating or editing the
$HOME/.hal/default/service-settings/<service>.yml of the respective service
By default, the following is set for all JVM-based services and provides 1/2 of the container’s allocated memory to be used by the service:
JAVA_OPTS=-XX:+UnlockExperimentalVMOptions -XX:+UseCGroupMemoryLimitForHeap -XX:MaxRAMFraction=2
(Note: The above is currently not being set for Gate and should therefore be set manually by the user)
You may find it useful to set Xmx and Xms of a JVM-based service. The example below shows how this would be done to the
gate service by editing/creating
env: JAVA_OPTS: "-Xms512m -Xmx1024m"
Make sure to
hal deploy apply for your changes to take effect.