Commits on Oct 18, 2021

1. Change CRD retry-backoff …

   Previously, if a custom resource failed to sync with Consul, we used the
   default retry backoff. This was an exponential backoff starting at 5ms
   and maxing out at 1000s (16m).
   
   This backoff was a poor UX for our common error case where one config
   entry cannot be applied due to a prerequisite, for example an ingress
   gateway entry cannot be applied until the protocol is set to http. The
   usual workflow to resolve this would be to look up the error, figure out
   the correct ServiceDefaults/ProxyDefaults, and then apply that resource.
   Once applied, the user needs to wait fo the ingress gateway (in this
   example) resource to be retried. With the default backoff config,
   because the user will have taken on the order of minutes to figure out
   the correct config, the exponential backoff will now be upwards of
   five minutes. The user will have to wait for a long time for the ingress
   gateway resource to be retried.
   
   This PR changes the backoff to start out at 200ms and max out at 5s.
   This fits our use-case better because the user will have to wait at max
   5ms, usually if there's an error, retrying within 5ms does nothing, so
   waiting 200ms to start is fine, and finally, Consul servers can accept
   tens of thousands of writes per second so even if there are a ton of
   resources retrying every 5s, it won't be an issue.

   

   lkysow committed Oct 18, 2021
   Configuration menu

   • View commit details
   • Copy full SHA for 7089f16
   • Browse repository at this point

   Copy the full SHA

   7089f16 View commit details

   Browse the repository at this point in the history