Closed Quentin-M closed 6 years ago
I'll be happy to make that a variable with a default. However note that until
will retry until the command actually succeeds, so it's mostly only about how many time the echo
occurs at this point.
Nevermind, I overlooked the until.
Thank you so much!
In certain conditions, the network may not entirely ready when the container starts. This prevents the plugins from being downloaded, which may not be desirable as liveness probes won't pick up the fact that ES does not run all the features that were expected to be present (e.g. prometheus metrics exporting).
Instead, this commit proposes retries upon installation failures. In Kubernetes, the pod will eventually get killed and restarted by the liveness probs after the initialDelaySeconds delay has expired.
The following happens pretty frequently on k8s 1.9+ w/ Weave CNI & CoreDNS. This makes Prometheus and the Alert Manager unhappy, as they consider that the elastic-search cluster is out.