While rolling out xdsrelay in staging, we noticed difficulty in debugging a cache drift between upstream control plane and xdsrelay.
It was hard to associate errors in xdsrelay with the per service cache drift. It would be easy to track errors if we had granular per key stats.
While rolling out xdsrelay in staging, we noticed difficulty in debugging a cache drift between upstream control plane and xdsrelay. It was hard to associate errors in xdsrelay with the per service cache drift. It would be easy to track errors if we had granular per key stats.
Signed-off-by: Jyoti Mahapatra jmahapatra@lyft.com