News for hackers. Stuff that matters.
Re:FoundationDB vs. Jepsen by teraflop on Tuesday March 25, 2014 @03:01PM (#7468076)
That means a client submitted a write that was successfully committed, but the client never found out that it was successful. It's more or less inevitable in the face of network partitions.

You don't want the database to report that a transaction was successful before the data is actually committed, or else you lose consistency. But if you wait until after the commit happens, there's a chance that the client might no longer be reachable. There's no way around this aside from waiting an indefinite period of time for the partition to be resolved.

Re:FoundationDB vs. Jepsen by jrullmann on Tuesday March 25, 2014 @03:09PM (#7468155)
Right. Most of the time you would simply retry the transaction - that's what FoundationDB's automatic transaction retry loop does. The first two tests I ran do that, but the third (foundationdb-append-noretry) does not.
Re:FoundationDB vs. Jepsen by misframer on Tuesday March 25, 2014 @03:04PM (#7468103)
That makes perfect sense. If it is the case, I think it's worth mentioning that the acknowledge write happened when the partition happened. I assumed that it happened when everything was going well, which is odd.
Re:FoundationDB vs. Jepsen by jrullmann on Tuesday March 25, 2014 @03:17PM (#7468221)
Good idea - I've updated the blog post. Thanks for the feedback!