Re: How to reproduce serialization failure for a read only transaction. - Mailing list pgsql-hackers

From Florian Pflug
Subject Re: How to reproduce serialization failure for a read only transaction.
Date
Msg-id 2EF4D8F7-4973-446C-915A-71D2C97E58F9@phlo.org
Whole thread Raw
In response to Re: Re: How to reproduce serialization failure for a read only transaction.  (Jim Nasby <jim@nasby.net>)
List pgsql-hackers
On Jan7, 2014, at 00:38 , Jim Nasby <jim@nasby.net> wrote:
> This email and the previous one are an awesome bit of information,
> can we add it to the docs somehow? Even if it's just dumping the
> emails into a wiki page and referencing it?

Most of what I wrote there can be found in README-SSE, I think,
under "Apparent Serial Order of Execution", "Heap locking" and
"Index AM implementations".

I guess it'd be nice if we explained these things in the docs
somewhere, though I'm not sure what level of detail would be
appropriate. Maybe a good compromise would be to explain dependency
graphs, but skip over the different kinds of dependencies (ww, rw, wr).
Instead we could say that whenever a transaction *does see* another
transaction's modifications it must appear after that transaction in any
serial schedule, and whenever a transaction *might see* another
transaction's modifications but doesn't due to begin/commit ordering
it must appear before that transaction.

best regards,
Florian Pflug




pgsql-hackers by date:

Previous
From: Emre Hasegeli
Date:
Subject: Failed assertion root->hasLateralRTEs on initsplan.c
Next
From: Dimitri Fontaine
Date:
Subject: Re: Fixing pg_basebackup with tablespaces found in $PGDATA