Front50's health check fails if Redis is not running locally

[gkuchta]

Issue

Front50's health check fails if Redis is not installed/running locally. It appears that this is due to springboot's org.springframework.boot.actuate.health.RedisHealthIndicator bubbling up to the surface.

Expected behavior

With "spinnaker.redis.enabled" : false and Front50 configured to use S3 for persistence, we had expected that http://localhost:8080/health would return UP if the S3 check passed. Instead, it returned DOWN, despite S3 working correctly. (ie: we could create pipelines, applications, etc, and see them populating the S3 bucket)

Steps to reproduce

• Tell Front50 NOT to use Redis.
• Stop the Redis instance running locally.
• curl http://localhost:8080/health -- it will report DOWN
• Re-start Redis
• curl http://localhost:8080/health -- it will report UP
• netstat -atnp | grep 6379 will show connections from the Front50 JVM to Redis.

Remediation

We were able to disable the springboot health check in front50-local.yml, though it might be useful to add this to the default configuration.

management:
  health:
    redis:
      enabled: false

[TheDigitalNinja]

I experienced the exact same issue today and can confirm that the above config change worked for us.

[iandelahorne]

I experienced the same issue, and the above fix helped. It doesn't return UP immediately after boot which can be confusing while debugging. Looking at S3Support.java:100 this seems to be 45 seconds due to the timeout.

[ttomsu]

Just got bit by this too. @gregturn, can you but a conditional on this?

[ttomsu]

Thanks, @gregturn !

spinnaker/front50#174