Re: Why we lost Uber as a user

On Tue, Aug  2, 2016 at 07:30:22PM -0700, Alfred Perlstein wrote:
> So for instance, let's say there is a bug in the master's write to disk.
> The logical replication acts as a barrier from that bad write going to the
> slaves.   With bad writes going to slaves then any corruption experienced on
> the master will quickly reach the slaves and they too will be corrupted.
> 
> With logical replication a bug may be stopped at the replication layer.  At
> that point you can resync the slave from the master.
> 
> Now in the case of physical replication all your base are belong to zuul and
> you are in a very bad state.
> 
> That said with logical replication, who's to say that if the statement is
> replicated to a slave that the slave won't experience the same bug and also
> corrupt itself.
> 
> We may be saying the same thing, but still there is something to be said for
> logical replication... also, didnt they show that logical replication was
> faster for some use cases at Uber?

I saw from the Uber article that they weren't going to per-row logical
replication but _statement_ replication, which is very hard to do
because typical SQL doesn't record what concurrent transactions
committed before a new statement's transaction snapshot is taken, and
doesn't record lock order for row updates blocked by concurrent activity
--- both of which affect the final result from the query.

So, for statement replication, it is not a question of whether the code
has bugs, but that the replay is not 100% possible in all cases, unless
you switch to some statement-row-lock hybrid ability.

-- 
  Bruce Momjian  <bruce@momjian.us>        http://momjian.us
  EnterpriseDB                             http://enterprisedb.com

+ As you are, so once was I. As I am, so you will be. +
+                     Ancient Roman grave inscription +