On guaranteeing write quorum

[Jack Vanlightly <ja...@gmail.com>]

Hi all,

I've been exploring what would be required to make write quorum a guarantee
rather than best effort, in terms of the core synchronous replication
protocol.

I blogged about the protocol recently and described situations where a
ledger can end up with AQ replicated blocks of entries in the middle of a
ledger. This isn't known but many users so I wanted to shed some light on
that aspect of the protocol. You can read that here:
https://medium.com/splunk-maas/apache-bookkeeper-insights-part-1-external-consensus-and-dynamic-membership-c259f388da21

I have documented my findings on the changes to the core protocol that
would be required to ensure AQ only blocks do not get left in ledgers. You
can read that write-up here:
https://jack-vanlightly.com/blog/2021/12/7/tweaking-the-bookkeeper-protocol-guaranteeing-write-quorum

I'd be interested in a general discussion on the topic.

Best regards
Jack