If the run harness is unable to start the data services for an app instance, it will delete the persistent volumes for that instance and retry. This will force a reload of the data, which usually fixes errors that prevent instance from starting.
Reduce the time spent waiting for services to start for non-backend services. The current delay is necessary for warming large numbers of app servers, but is too long for the data services.
When building the docker images, only build the executables if the image(s) being built will need them. This speeds building just the run harness.
I tested this on a number of runs and it worked properly to catch and recover from a data corruption issue.
This pull request has the following changes:
I tested this on a number of runs and it worked properly to catch and recover from a data corruption issue.