News for hackers. Stuff that matters.
Re:Jepsen: Distributed Systems Safety Analysis by JoelJacobson on Wednesday October 21, 2015 @12:29AM (#10423724)
Why isn't PostgreSQL in the list on the first page? It is in the [blog posts] page though.

I would love to see the multixact data corruption problems introduced in 9.3 analyzed, and see if he can verify them to be solved in the latest version.

Re:Jepsen: Distributed Systems Safety Analysis by rdtsc on Wednesday October 21, 2015 @12:52AM (#10423788)
(Not author) but from what I remember PostreSQL was a bit of a funny case. It was about partitions between clients and a single database server. In most other cases, it was analysis of distributed back-end server instances. Not that those failures are not important but it is just harder to compare it seems with the others.
Re:Jepsen: Distributed Systems Safety Analysis by lomnakkus on Wednesday October 21, 2015 @03:15AM (#10424083)
Yeah, basically with PostgreSQL it boils down to the fact that even if a client receives an error on commit, its writes may in fact have succeeded -- and it cannot assume that they haven't. (Since the final "ack" that everything went well may not make it.)

As an example: This means that idempotency is really important if you're doing any kind of retries on commit failure.

EDIT: Another example: If you're doing application-level optimistic concurrency control (using e.g. an incrementing version number) you can end up in a situation where an update appears to conflict with itself. Usually, though, this isn't a big concern since client<->server paritions are usually catastrophic anyway.

(Off the top of my head that's what I can recall -- there may be other things to be gleaned from the original article.)

[1] At least for all practical purposes.

Re:Jepsen: Distributed Systems Safety Analysis by biokoda on Wednesday October 21, 2015 @03:39AM (#10424130)
> Yeah, basically with PostgreSQL it boils down to the fact that even if a client receives an error on commit, its writes may in fact have succeeded -- and it cannot assume that they haven't.

This is true for almost every database. Otherwise they would require a distributed transaction between the client and server.

Re:Jepsen: Distributed Systems Safety Analysis by lomnakkus on Friday October 23, 2015 @04:09PM (#10441014)
Indeed, but given that we were talking about $DISTSYS vs. $CONCRETE_THING, I just mentioned the $CONCRETE_THING by name.

You are of course technically right... which is the best kind of right! :)

Re:Jepsen: Distributed Systems Safety Analysis by amitlan on Wednesday October 21, 2015 @02:19AM (#10423980)
Not quite sure whether something like multixact data corruption is symptomatic of the kind of underlying issues in database system implementations that Jepsen is after. I may be wrong though.
Re:Jepsen: Distributed Systems Safety Analysis by Jweb_Guru on Wednesday October 21, 2015 @11:03AM (#10425898)
It wasn't. It's also rather difficult to trigger. I have no doubt that aphyr could do stuff like that if he so chose, but I think pounding on getting a reproducible test case for Postgres's subtle serializable bug would be more his speed.
Re:Jepsen: Distributed Systems Safety Analysis by threeseed on Wednesday October 21, 2015 @03:46AM (#10424157)
PostgreSQL was tested in single mode. All others distributed.

Apples and Oranges.