To make secure proxy perform better, we need to remove the hard coded max connection value as to allow more connections when using larger VM sizes. We also need to add a healthcheck/drain piece to the openresty config to support CF Drain
Review NGINX best practices and implement any other performance related config items and make defaults as well as config options
Create lua health checks to follow CF gorouter drain setup
gorouter has a drain_wait value which defaults to 20 seconds - during that time, the gorouter health endpoint will give 503s but still accept and route HTTP traffic on it's main port during the drain_wait. This delay/wait allows the ELB health check to kick the server out of the pool. Currently secure proxy downs both it's health check proxy endpoint and the gorouter proxy port once the gorouter health check throws it's first 503 - this causes 502s to customers until the ELB health checks remove the VM from the target pool
Update and test any packages in the release to latest/support versions
To make secure proxy perform better, we need to remove the hard coded max connection value as to allow more connections when using larger VM sizes. We also need to add a healthcheck/drain piece to the openresty config to support CF Drain
Notes
drain_wait
value which defaults to 20 seconds - during that time, the gorouter health endpoint will give 503s but still accept and route HTTP traffic on it's main port during thedrain_wait
. This delay/wait allows the ELB health check to kick the server out of the pool. Currently secure proxy downs both it's health check proxy endpoint and the gorouter proxy port once the gorouter health check throws it's first 503 - this causes 502s to customers until the ELB health checks remove the VM from the target poolAcceptance Criteria