Anyway, moving on: I'd like to hear from y'all about what kind of Postgres replication you think is safest. Patroni, perhaps?
-
Show this thread
-
"Note: Because of the way synchronous replication is implemented in PostgreSQL it is still possible to lose transactions even when using synchronous_mode_strict" oh, okay, cool is there a postgres replication/HA setup that *doesn't* lose data? https://github.com/zalando/patroni/blob/master/docs/replication_modes.rst#synchronous-mode …
7 replies 5 retweets 24 likesShow this thread -
Replying to @jepsen_io
This is the behaviour of PostgreSQL and there is no good and reliable way to overcome it.
1 reply 0 retweets 0 likes -
Replying to @cyberdemn
I believe you! What about... commercial offerings? Vertica, perhaps?
1 reply 0 retweets 0 likes -
Replying to @jepsen_io
Actually I would be really interested in testing Patroni. It is a good opportunity to make it better :)
1 reply 0 retweets 2 likes -
Replying to @cyberdemn
I'd be happy to do it, but if there's already well-documented behavior around lost transactions, I'm not sure if any of Jepsen's more sophisticated testing will be super useful--unless it's possible to rule out that data loss somehow?
2 replies 0 retweets 1 like -
Replying to @jepsen_io @cyberdemn
Stolon's docs, meanwhile, make this interesting claim around sync replication txn loss: "with stolon we have the power to overcome this issue". Maybe that's worth putting to the test? https://github.com/sorintlab/stolon/blob/master/doc/syncrepl.md#handling-postgresql-sync-repl-limits-under-such-circumstances …
1 reply 0 retweets 1 like -
Replying to @jepsen_io
It solves it only partially, when the primary is restarted. There are other ways to get into trouble. It is enough just to kill the client connection.
1 reply 0 retweets 2 likes -
-
Replying to @cyberdemn @jepsen_io
Isn't it a multi failure scenario? How many consecutive/simultaneous failures the clusterware should handle? This documented failure involve a primary or client failure, THEN a primary crash BEFORE it could replicate, THEN a standby election.
1 reply 0 retweets 0 likes
I mean, I'd argue that a database which is paying a round-trip latency penalty on every write can and should be getting consensus for that cost, rather than losing data!
-
-
Replying to @jepsen_io @cyberdemn
indeed. There is a (documented) hole in the consensus with this corner case :( Interesting issue.
0 replies 0 retweets 0 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.