markusthoemmes commented on issue #2859: LoadBalancerService does not consider concurrent requests when scheduling URL: https://github.com/apache/incubator-openwhisk/issues/2859#issuecomment-337065359 This is not necessarily related to the shared state, it is also the case with the local version of the load balancer. Although that used AtomicIntegers, they were simply used as a thread-safe counter in terms of changing operations, but not used as an atomically checked-then-update counter. Should we implement the Loadbalancer as an Actor to achieve local consistency for each Controller to start? That would solve parallel read consistency at least. For the write/read races introduced because of state-sharing, we can come up with a smart solution as outlined by @tysonnorris to keep local state synchronous. Maybe there's a mechanism to wait for a write to happen and then implement the respective operations as `ask`s so they "block" scheduling until that returned? As noted by @vvraskin both of these bits would need to be tested under load but I'm confident that at least the "load balancer as an Actor" should not impose a bottleneck. ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
With regards, Apache Git Services
