We have observed an issue in Coopr, which we expect to also exist in CDAP, since in both cases the endpoints are built on top of netty-http. The Coopr issue is https://issues.cask.co/browse/COOPR-622, which has more details.
The issue is that the /status endpoint is not closing the connection, which in the case of Google Compute and likely other providers, fails the load-balancer health-checks.
We have observed an issue in Coopr, which we expect to also exist in CDAP, since in both cases the endpoints are built on top of netty-http. The Coopr issue is https://issues.cask.co/browse/COOPR-622, which has more details.
The issue is that the /status endpoint is not closing the connection, which in the case of Google Compute and likely other providers, fails the load-balancer health-checks.