EXPLAIN BUFFERS: dirtied - Mailing list pgsql-general

From Vitaliy Garnashevich
Subject EXPLAIN BUFFERS: dirtied
Date
Msg-id 21a14f5f-b39c-3aa8-da2c-2112ee328b93@gmail.com
Whole thread Raw
Responses Re: EXPLAIN BUFFERS: dirtied  (Tom Lane <tgl@sss.pgh.pa.us>)
List pgsql-general
Hi,

In EXPLAIN (ANALYZE, BUFFERS) for a SELECT query, I see the following 
statistics under an Index Scan node:

Buffers: shared hit=8357288 read=6165444 dirtied=44820 written=5590

As far as I understand, that's the statistics for accesses to shared 
buffers during the query:
- hit = required page was already in shared buffers
- read = required page was not in shared buffers, and was loaded from 
disk (from filesystem cache)
- written = while loading the required page, there was no free space for 
it in shared buffers, so some other dirty page was evicted from shared 
buffers and was written to disk (to filesystem cache), to free some 
space and to load the required page

But what is "dirtied" statistics? When a SELECT query could make pages 
dirty?

Regards,
Vitaliy



pgsql-general by date:

Previous
From: Pritam Barhate
Date:
Subject: Re: Using AWS ephemeral SSD storage for production database workload?
Next
From: Tom Lane
Date:
Subject: Re: EXPLAIN BUFFERS: dirtied