Hot Standby (v9d) - Mailing list pgsql-hackers

From Simon Riggs
Subject Hot Standby (v9d)
Date
Msg-id 1232663710.2327.1028.camel@ebony.2ndQuadrant
Whole thread Raw
Responses Re: Hot Standby (v9d)  (Heikki Linnakangas <heikki.linnakangas@enterprisedb.com>)
Re: Hot Standby (v9d)  (Simon Riggs <simon@2ndQuadrant.com>)
List pgsql-hackers
Latest version of Hot Standby includes new deferred cancellation for
conflict resolution and further refactoring.

http://wiki.postgresql.org/wiki/Hot_Standby#Usage

[Please note that max_standby_delay parameter has been set to default to
zero (0), to allow investigation of usability. i.e. queries will be
cancelled fairly quickly if they conflict. Please set this in
recovery.conf to any value you consider reasonable and report findings.]

GENERAL PLANS

* Bug fix v9 over next few days
* Put corrected version into GIT
* Produce outstanding items as patch-on-patch via GIT
* Work with reviewers to nibble away until fully committed

OUTSTANDING ITEMS (as of Version 9d)

Reported problems (Identified by) (Target release)

* None (as yet!) - please post all bugs to list
All v9 bugs have been fixed. This version has passed initial bash and
function tests and we will continue testing this version tomorrow.

Main work/required additional items (Proposed by) (Target release)

      * Complete Prepared Transaction support (Simon)

Performance tuning opportunities (Proposed by) (Target release)

      * tuning of RecordKnownAssignedTransactionIds() using hash table
        (Heikki)

Usability enhancements

* don't cancel all currently connected users every time we drop a
tablespace that has temp files being used within that tablespace -
should be able to parse the filenames to get pids to cancel

* don't kill all currently connected users every time we drop a user -
should be able to match list of current users against connected roles

* more work possible on btree deletes to reduce impact - should be able
to work out a reasonable latestRemovedXid


Testing of any kind welcome, the heavier the better. Save the database
and logs if it crashes, please. Performance data especially valuable.

--
 Simon Riggs           www.2ndQuadrant.com
 PostgreSQL Training, Services and Support

Attachment

pgsql-hackers by date:

Previous
From: Tom Lane
Date:
Subject: Re: autovacuum, reloptions, and hardcoded pg_class tupdesc
Next
From: Josh Berkus
Date:
Subject: Re: Pluggable Indexes (was Re: rmgr hooks (v2))