2 19.9. Run-time Statistics #
4 19.9.1. Cumulative Query and Index Statistics
5 19.9.2. Statistics Monitoring
7 19.9.1. Cumulative Query and Index Statistics #
9 These parameters control the server-wide cumulative statistics system.
10 When enabled, the data that is collected can be accessed via the
11 pg_stat and pg_statio family of system views. Refer to Chapter 27 for
14 track_activities (boolean) #
15 Enables the collection of information on the currently executing
16 command of each session, along with its identifier and the time
17 when that command began execution. This parameter is on by
18 default. Note that even when enabled, this information is only
19 visible to superusers, roles with privileges of the
20 pg_read_all_stats role and the user owning the sessions being
21 reported on (including sessions belonging to a role they have
22 the privileges of), so it should not represent a security risk.
23 Only superusers and users with the appropriate SET privilege can
26 track_activity_query_size (integer) #
27 Specifies the amount of memory reserved to store the text of the
28 currently executing command for each active session, for the
29 pg_stat_activity.query field. If this value is specified without
30 units, it is taken as bytes. The default value is 1024 bytes.
31 This parameter can only be set at server start.
33 track_counts (boolean) #
34 Enables collection of statistics on database activity. This
35 parameter is on by default, because the autovacuum daemon needs
36 the collected information. Only superusers and users with the
37 appropriate SET privilege can change this setting.
39 track_cost_delay_timing (boolean) #
40 Enables timing of cost-based vacuum delay (see Section 19.10.2).
41 This parameter is off by default, as it will repeatedly query
42 the operating system for the current time, which may cause
43 significant overhead on some platforms. You can use the
44 pg_test_timing tool to measure the overhead of timing on your
45 system. Cost-based vacuum delay timing information is displayed
46 in pg_stat_progress_vacuum, pg_stat_progress_analyze, in the
47 output of VACUUM and ANALYZE when the VERBOSE option is used,
48 and by autovacuum for auto-vacuums and auto-analyzes when
49 log_autovacuum_min_duration is set. Only superusers and users
50 with the appropriate SET privilege can change this setting.
52 track_io_timing (boolean) #
53 Enables timing of database I/O waits. This parameter is off by
54 default, as it will repeatedly query the operating system for
55 the current time, which may cause significant overhead on some
56 platforms. You can use the pg_test_timing tool to measure the
57 overhead of timing on your system. I/O timing information is
58 displayed in pg_stat_database, pg_stat_io (if object is not
59 wal), in the output of the pg_stat_get_backend_io() function (if
60 object is not wal), in the output of EXPLAIN when the BUFFERS
61 option is used, in the output of VACUUM when the VERBOSE option
62 is used, by autovacuum for auto-vacuums and auto-analyzes, when
63 log_autovacuum_min_duration is set and by pg_stat_statements.
64 Only superusers and users with the appropriate SET privilege can
67 track_wal_io_timing (boolean) #
68 Enables timing of WAL I/O waits. This parameter is off by
69 default, as it will repeatedly query the operating system for
70 the current time, which may cause significant overhead on some
71 platforms. You can use the pg_test_timing tool to measure the
72 overhead of timing on your system. I/O timing information is
73 displayed in pg_stat_io for the object wal and in the output of
74 the pg_stat_get_backend_io() function for the object wal. Only
75 superusers and users with the appropriate SET privilege can
78 track_functions (enum) #
79 Enables tracking of function call counts and time used. Specify
80 pl to track only procedural-language functions, all to also
81 track SQL and C language functions. The default is none, which
82 disables function statistics tracking. Only superusers and users
83 with the appropriate SET privilege can change this setting.
87 SQL-language functions that are simple enough to be “inlined”
88 into the calling query will not be tracked, regardless of this
91 stats_fetch_consistency (enum) #
92 Determines the behavior when cumulative statistics are accessed
93 multiple times within a transaction. When set to none, each
94 access re-fetches counters from shared memory. When set to
95 cache, the first access to statistics for an object caches those
96 statistics until the end of the transaction unless
97 pg_stat_clear_snapshot() is called. When set to snapshot, the
98 first statistics access caches all statistics accessible in the
99 current database, until the end of the transaction unless
100 pg_stat_clear_snapshot() is called. Changing this parameter in a
101 transaction discards the statistics snapshot. The default is
106 none is most suitable for monitoring systems. If values are only
107 accessed once, it is the most efficient. cache ensures repeat
108 accesses yield the same values, which is important for queries
109 involving e.g. self-joins. snapshot can be useful when
110 interactively inspecting statistics, but has higher overhead,
111 particularly if many database objects exist.
113 19.9.2. Statistics Monitoring #
115 compute_query_id (enum) #
116 Enables in-core computation of a query identifier. Query
117 identifiers can be displayed in the pg_stat_activity view, using
118 EXPLAIN, or emitted in the log if configured via the
119 log_line_prefix parameter. The pg_stat_statements extension also
120 requires a query identifier to be computed. Note that an
121 external module can alternatively be used if the in-core query
122 identifier computation method is not acceptable. In this case,
123 in-core computation must be always disabled. Valid values are
124 off (always disabled), on (always enabled), auto, which lets
125 modules such as pg_stat_statements automatically enable it, and
126 regress which has the same effect as auto, except that the query
127 identifier is not shown in the EXPLAIN output in order to
128 facilitate automated regression testing. The default is auto.
132 To ensure that only one query identifier is calculated and
133 displayed, extensions that calculate query identifiers should
134 throw an error if a query identifier has already been computed.
136 log_statement_stats (boolean)
137 log_parser_stats (boolean)
138 log_planner_stats (boolean)
139 log_executor_stats (boolean) #
140 For each query, output performance statistics of the respective
141 module to the server log. This is a crude profiling instrument,
142 similar to the Unix getrusage() operating system facility.
143 log_statement_stats reports total statement statistics, while
144 the others report per-module statistics. log_statement_stats
145 cannot be enabled together with any of the per-module options.
146 All of these options are disabled by default. Only superusers
147 and users with the appropriate SET privilege can change these