Re: 16-bit page checksums for 9.2

From: "Kevin Grittner" <Kevin(dot)Grittner(at)wicourts(dot)gov>
To: <simon(at)2ndQuadrant(dot)com>
Cc: <heikki(dot)linnakangas(at)enterprisedb(dot)com>,<aidan(at)highrise(dot)ca>, <stark(at)mit(dot)edu>, <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: 16-bit page checksums for 9.2
Date: 2011-12-30 14:33:15
Message-ID: 4EFD774B0200002500044242@gw.wicourts.gov
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

> Simon Riggs wrote:
> Kevin Grittner wrote:

>> if there is no checksum in the page itself, you can put one in the
>> double-write metadata.

> However, I don't see that it provides protection across non-crash
> write problems. We know we have these since many systems have run
> without a crash for years and yet still experience corrupt data.

Agreed. I don't think anyone has tried to assert it solves the same
problems that checksums solve -- it is a high-performance way to
solve some of the problems that an in-page checksum *creates* without
breaking pg_upgrade.

> Double writes do not require page checksums but neither do they
> replace page checksums.

To nit-pick: double writes require a page checksum, but (as Heikki
pointed out) they don't require it to be stored in the page. If
there *is* one stored in the page, it probably makes sense to use it.

> So I think we need page checksums plus either FPWs or double
> writes.

Adding checksums by themselves creates a risk of false positive
corrupted page indications following an OS or hardware crash.
Additional FPWs or a new double-write mechanism are two of miriad
possible solutions to that. If it is going to be addressed for 9.2,
I believe they're the two most reasonable, especially from the POV of
pg_upgrade.

So, while they should be separate patches, the complement each other;
each makes the other perform better, and they should share some code.

-Kevin

Browse pgsql-hackers by date

  From Date Subject
Next Message Aidan Van Dyk 2011-12-30 14:44:09 Re: 16-bit page checksums for 9.2
Previous Message Kevin Grittner 2011-12-30 14:18:43 Re: 16-bit page checksums for 9.2