postgresql/contrib/pg_autovacuum

pg_autovacuum README
--------------------

pg_autovacuum is a libpq client program that monitors all the
databases associated with a PostgreSQL server.  It uses the statistics
collector to monitor insert, update and delete activity.  

When a table exceeds a insert or delete threshold (for more detail on
thresholds, see "Vacuum and Analyze" below) then that table will be
vacuumed and/or analyzed.

This allows PostgreSQL to keep the FSM (Free Space Map) and table
statistics up to date, and eliminates the need to schedule periodic
vacuums.

The primary benefit of pg_autovacuum is that the FSM and table
statistic information are updated more nearly as frequently as needed.
When a table is actively changing, pg_autovacuum will perform the
VACUUMs and ANALYZEs that such a table needs, whereas if a table
remains static, no cycles will be wasted performing this
unnecessarily.

A secondary benefit of pg_autovacuum is that it ensures that a
database wide vacuum is performed prior to XID wraparound.  This is an
important, if rare, problem, as failing to do so can result in major
data loss.  (See the section in the _Administrator's Guide_ entitled
"Preventing transaction ID wraparound failures" for more details.)

KNOWN ISSUES:
-------------

pg_autovacuum has been tested under Redhat Linux (by me) and Debian
GNU/Linux, Solaris, and AIX (by Christopher B. Browne) and all known
bugs have been resolved.  Please report any problems to the hackers
list.

pg_autovacuum requires that the statistics system be enabled and
reporting row level stats.  The overhead of the stats system has been
shown to be significant under certain workloads.  For instance, a
tight loop of queries performing "select 1" was found to run nearly
30% slower when row-level stats were enabled.  However, in practice,
with more realistic workloads, the stats system overhead is usually
nominal.

pg_autovacuum does not get started automatically by either the
postmaster or by pg_ctl.  Similarly, when the postmaster exits, no one
tells pg_autovacuum.  The result of that is that at the start of the
next loop, pg_autovacuum will fail to connect to the server and
exit().  Any time it fails to connect pg_autovacuum exit()s.

While pg_autovacuum can manage vacuums for as many databases as you
may have tied to a particular PostgreSQL postmaster, it can only
connect to a single PostgreSQL postmaster.  Thus, if you have multiple
postmasters on a particular host, you will need multiple pg_autovacuum
instances, and they have no way, at present, to coordinate between one
another to ensure that they do not concurrently vacuum big tables.

When installed as a service under Windows, there is currently no way to
know the name of the PostgreSQL server service (if there even is one)
so it is not possible to specify a startup dependency. It is therefore
possible for pg_autovacuum to start before the server.

When installed as a service under Windows, if the -P option is used to
specify the connection password, this option (and the password) is 
stored in plain text in the registry.

TODO:
-----

At present, there are no sample scripts to automatically start up
pg_autovacuum along with the database.  It would be desirable to have
a SysV script to start up pg_autovacuum after PostgreSQL has been
started.

Some users have expressed interest in making pg_autovacuum more
configurable so that certain tables known to be inactive could be
excluded from being vacuumed.  It would probably make sense to
introduce this sort of functionality by providing arguments to specify
the database and schema in which to find a configuration table.

It would also be desirable for the daemon to monitor how busy the
system is, with a view to deferring vacuums until there is less other
activity.

INSTALL:
--------

As of postgresql v7.4 pg_autovacuum is included in the main source
tree under contrib.  Therefore you merely need to "make && make
install" (similar to most other contrib modules) and it will be
installed for you.

If you are using an earlier version of PostgreSQL, uncompress the
tar.gz file into the contrib directory and modify the contrib/Makefile
to include the pg_autovacuum directory.  pg_autovacuum will then be
built as part of the standard postgresql install.  It is known to work
with v7.3 releases; it is not presently compatible with v7.2.

make sure that the following are set in postgresql.conf:

  stats_start_collector = true
  stats_row_level = true

Start up the postmaster, then execute the pg_autovacuum executable.

If you have a script that automatically starts up the PostgreSQL
instance, you might add in, after that, something similar to the
following:

  sleep 10    # To give the database some time to start up
  $PGBINS/pg_autovacuum -D -s $SBASE -S $SSCALE ... [other arguments]

Command line arguments:
-----------------------

pg_autovacuum has the following optional arguments:

-d debug: 0 silent, 1 basic info, 2 more debug info,  etc...
-D daemonize: Detach from tty and run in background.
-s sleep base value: see "Sleeping" below.
-S sleep scaling factor: see "Sleeping" below.
-v vacuum base threshold: see "Vacuum and Analyze" below.
-V vacuum scaling factor: see "Vacuum and Analyze" below.
-a analyze base threshold: see "Vacuum and Analyze" below.
-A analyze scaling factor: see "Vacuum and Analyze" below.
-i update interval: how often (in terms of iterations of the primary loop
   over the database list) to update the database list. The default is 2,
   which means the list will be updated before every other pass through
   the database list.
-L log file: Name of file to which output is submitted, otherwise STDERR
-U username: Username pg_autovacuum will use to connect with, if not
   specified the current username is used.
-P password: Password pg_autovacuum will use to connect with. *WARNING*
   This option is insecure. When installed as a Windows Service, this
   option will be stored in plain text in the registry. When used with
   most Unix variants, other users will be able to see the argument to
   the "-P" option via ps(1). The ~/.pgpass file can be used to
   specify a password more securely.
-H host: host name or IP to connect to.
-p port: port used for connection.
-h help: list of command line options.

The following 5 autovacuum command line options correspond to the various
cost-based vacuum settings.  If not given, then the cluster default values
will be used.

-c	vacuum_cost_delay
-C	vacuum_cost_page_hit
-m	vacuum_cost_page_miss
-n	vacuum_cost_page_dirty
-l	vacuum_cost_limit


Numerous arguments have default values defined in pg_autovacuum.h.  At
the time of writing they are:

-d 1
-v 1000
-V 2   
-a 500 (half of -v if not specified)
-A 1   (half of -V if not specified)
-s 300 (5 minutes)
-S 2
-i 2

The following arguments are used on Windows only:

-I Install the executable as a Windows service. Other appropriate command
   line options will be stored in the registry and passed to the service 
   at startup. *WARNING* This includes the connection password which will 
   be stored in plain text.
	     
-N service user: Name of the Windows user account under which the service
   will run. Only used when installing as a Windows service.
   
-W service password: The password for the service account. Only used when 
   installing as a Windows service.

-R Uninstall pg_autovacuum as a service.

-E Dependent service that must start before this service. Normally this will be
   a PostgreSQL instance, e.g. "-E pgsql-8.0.0". Only used when installing as 
   a Windows service.

Vacuum and Analyze:
-------------------

pg_autovacuum performs either a VACUUM ANALYZE or just ANALYZE
depending on the mixture of table activity (insert, update, or
delete):

- If the number of (inserts + updates + deletes) > AnalyzeThreshold, then
  only an analyze is performed.

- If the number of (deletes + updates) > VacuumThreshold, then a
  vacuum analyze is performed.

VacuumThreshold is equal to:
    vacuum_base_value + (vacuum_scaling_factor * "number of tuples in the table")

AnalyzeThreshold is equal to:
    analyze_base_value + (analyze_scaling_factor * "number of tuples in the table")

The AnalyzeThreshold defaults to half of the VacuumThreshold since it
represents a much less expensive operation (approx 5%-10% of vacuum),
and running ANALYZE more often should not substantially degrade system
performance.

Sleeping:
---------

pg_autovacuum sleeps for a while after it is done checking all the
databases.  It does this in order to limit the amount of system
resources it consumes.  This allows the system administrator to
configure pg_autovacuum to be more or less aggressive.

Reducing the sleep time will cause pg_autovacuum to respond more
quickly to changes, whether they be database addition/removal, table
addition/removal, or just normal table activity.

On the other hand, setting pg_autovacuum to sleep values too
aggressively (to too short periods of time) can have a negative effect
on server performance.  For instance, if a table gets vacuumed 5 times
during the course of a large set of updates, this is likely to take a
lot more work than if the table was vacuumed just once, at the end.

The total time it sleeps is equal to:

  base_sleep_value + sleep_scaling_factor * "duration of the previous
  loop"

Note that timing measurements are made in seconds; specifying
"pg_vacuum -s 1" means pg_autovacuum could poll the database up to 60
times minute.  In a system with large tables where vacuums may run for
several minutes, rather longer times between vacuums are likely to be
appropriate.

What pg_autovacuum monitors:
----------------------------

pg_autovacuum dynamically generates a list of all databases and tables
that exist on the server.  It will dynamically add and remove
databases and tables that are removed from the database server while
pg_autovacuum is running.  Overhead is fairly small per object.  For
example: 10 databases with 10 tables each appears to less than 10k of
memory on my Linux box.