Agenda:

  1. Status Updates
    1. 4.8, release branch to be created soon (Enrico)
    1. Proposals for using Binary format in METADATA  - status of the work 
    2. Immutable Ledger Metadata - status of the work  (Sijie)
    3. New "Duracility Contract Checker" (Charan)
    4. Documentation status (need to document new features in 4.8)
    5. ByteBuf ownership (Andrey)
    6. ExplicitLAC on the reader side (Enrico)
  2. BPs https://github.com/apache/bookkeeper/pulls?utf8=%E2%9C%93&q=is%3Aopen+bp+
  3. PRs https://github.com/apache/bookkeeper/pulls

 

Attendees:

JV, Enrico, Sijie, Andrey, Charan


Minutes:

  • JV askes to Sijie if there are known cases of error, in which a ledger comes back to OPEN state from CLOSE state
  • Sijie: Ivan is refactoing LedgerHandle in order to  not to have dependencies to "BookKeeper" class, so that it will be more 'testable'
  • Enrico: all of the changes are huge, I it preferrable to delay them to 4.9, Sijie is okay. Let's start and email thread on the mailing list
  • JV asks more explanations about the usage of watchers on the reader side as Enrico was discussing about, Sijie remembers that he introduced it at Twitter, but there are cases in which it makes sense not to have a watch, which is a waste, we can re-read metadata in case of missing information
  • Charan is wokring on a "Durability Constract", that is a new checker which verifies that all the data is replicated to enough bookies
    • a new RPC will be added to ask to a Bookie the list of entry ids that are present on the Bookie, it is important that such implementation won't need to scan the bookie, but this information must be obtained from index
  • JV: We talked multiple times about changing from 63bit ledger id to 128bit / UUID, which is the real problem ? 
    • Sijie: it would be better to start an email thread, so that we can recap and restart that thread
  • No labels