OPA startup failure can slow down routing table creation

[szuecs]

OPA startup failure can slow down routing table creation

Route table update latency spikes:

Logs:

Starting OPA instance..." bundle-name=my-app
Closing OPA instance..." bundle-name=my-app
failed to process route kube_rg__routeID: invalid_filter_params: failed to create filter \"opaServeResponse\": open policy agent instance for bundle name 'my-app' could not be obtained: failed to start OPA instance for bundle \"my-app\": one or more open policy agent plugins failed to start in 30s with error: timed out while starting: context deadline exceeded" 

[Pushpalanka]

If this is on route update scenario,
Enabling pre-loading will support here. Then route table will be updated without waiting for OPA instance to start and become healthy. During that time route will be serving 503 until OPA becomes healthy, but route table update will go through without blocking on OPA start.

Is this about the Skipper start up itself or a case where pre-loading can't be considered due to some reason?

[szuecs]

I have to check if preloading is enabled
I saw only 3 lines of logs per pod so my assumption is that it was only at startup.