[GitHub] tysonnorris commented on issue #2859: LoadBalancerService does not consider concurrent requests when scheduling

[gi...@git.apache.org]

tysonnorris commented on issue #2859: LoadBalancerService does not consider concurrent requests when scheduling
URL: https://github.com/apache/incubator-openwhisk/issues/2859#issuecomment-337057445
 
 
   I think "reading your own writes" is only for the replicated data, which is inherently async (by way of messaging with replicator), so I'm not sure that will work (but great if it does!).
   
   I was thinking about the following:
   - revert to the LoadBalancerData to sync signatures (e.g. `def activationCountPerInvoker: Map[String, Int]`
   - make DistributedLoadBalancerData keep 2 sets of data - one local, one replicated
   - the views that DistributedLoadBalancerData exposes would be a composition of local + replicated data 
   
   Note that this requires changing the replicated format to include an indication of the controller instance that contributed the data, so that the local instance can be ignored within the set of replicated data, in favor of the more accurate local data. Basically the localdata would be accurate and the replicated data would be eventually consistent. But at least within the same controller the scheduling would be predictable under load. 
   
   
     
 
----------------------------------------------------------------
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:
users@infra.apache.org


With regards,
Apache Git Services