Hi, On 2016-04-01 08:46:01 +0200, Andres Freund wrote: > That's a fundamental misunderstanding on your part (perhaps created by > imprecise docs).
> > Speaking of which, did you see the proposed README I sent for > > src/backend/replication/logical ? > > I skimmed it. But given we have a CF full of patches, some submitted > over a year ago, it seems unfair to spend time on a patch submitted a > few days ago.
That's not up2date obviously, but it still might help.
Thanks, I've been reading it and the posts it references.
Most of it was familiar by this point, but would've been a good reference earlier on. The snapshot builder docs in README.SNAPBUILD.txt are handy and help glue a few separate pieces together better for me, and the invalidations section was brief but informative.
The very last point looks interesting, but only really alludes to what's going on:
+== Restartable Decoding ==
+
+As we want to generate a consistent stream of changes we need to have the
+ability to start from a previously decoded location without waiting possibly
+very long to reach consistency. For that reason we dump the current visibility
+information to disk whenever we read an xl_running_xacts record.
I don't feel like I've grasped this properly yet. I think it's referring to the pg_logical/snapshots/ serialization, the use of which allows us to avoid doing extra work in SnapBuildFindSnapshot(...), but doesn't seem to be crucial for correct function. After all, decoding still restarts at the restart_lsn and feeds relevant xact info into the snapshot builder, accumulates invalidation information, etc.
After 9.6 I'd like to go through that, update it, and get it in as a README for logical decoding. It would've done me a lot of good when getting up to speed.
