Closed Shivs11 closed 2 weeks ago
This PR is going to be marked as closed. Discussions with @ShahabT proved that merging AckManager into db, as was done here, was wrong since write latencies could occur. Moreover, moving ackLevel into db
also did not make sense as db
would then in-turn have to wait for AckManager's lock which could further cause latencies.
Thus, persisting the latest AckLevel with task creation, as a result, will also not be done at this moment in time.
What changed?
AckLevel
inside ofdb
from the AckManager. This is done so that we persist AckLevels with every batch of tasks that are created.readLevel
(amongst other variables) which would make life harder and code unclearer in the near future.db
altogether. This shall also be in agreement with our long term plan of merging code entirely intobacklogManager
in the near future.Why?
How did you test it?
Potential risks
None, merging into my own feature branch
Is hotfix candidate?
No