pg_monitor 1.7.3

PostgreSQL monitoring checks

# pg_monitor
Monitors postgres and produces output usable by monitoring systems such as Nagios

Checks for a PostgreSQL server

Installation :
pip install pg_monitor

Requirement :
python >= 2.7

Platform :
Windows, Linux

Bug report :
Please send your bug reports to [e-mail]

optional arguments:
-h, --help show this help message and exit
--port PORT Port of the database to be accessed. Default: 5432
--user USER Username allowed to access the database. Default:
postgres
--password PASSWORD Password of the supplied user. Default:
--dbname DBNAME The name of the database to be accessed. Default:
postgres
--host HOST Host of the database. Default: localhost
--check {backends,connections,wals,vacuum,autovacuum,analyze,autoanalyze,table_bloat,
index_bloat,blocking,nonblocking,table_size,index_size,database_size,
checkpoints,duplicate_indexes,replica_lag}

Options to be passed as checks :

** backends -->> Checks for the number of clients/connections currently
connected to the specified database
** connections -->> Tests the databases if they currently accept
connections or not
** wals -->> keeps tab on the number/size of wal files present in the pg_xlog
directory
** vacuum -->> Inspects the tables and produces a report of those that has not
been vauumed for a given period of time
** autovacuum -->> Same as vacuum except that this is triggered by the database's
internal processes
** analyze -->> reports tables that have not been analyzed after a given period
of time
** autoanalyze -->> Same as analyze but triggered by the database's internal
processes.
** table_bloat -->> Inspect all tables and reports the bloat percentage
which is usually a function of vacuum and/or autovacuuming not running regularly
** index_bloat -->> Same as table bloat, except for indexes. This could be
remedied by reindexing
** blocking -->> This checks for locks transactional/relational that are blocking
other processes from running.
** nonblocking -->> This checks for locks that are running more than the specified
type. This normally do not block, but could become a blocking lock if ir runs
longer than necessary. In some cases, it is the long running locks that become
the blocking types
** table_size -->> Monitors table sizes
** index_size -->> Monitors index sizes
** database_size -->> Monitors database sizes
** checkpoints -->> Monitors checkpoint frequency
** duplicate_indexes -->> Monitors indexes and report duplicates
** replica_lag -->> Monitors all replicas connected to a given master
--find FIND [FIND ...] Finding partterns in locks
--warning WARNING Values to be considered as warning signs
--critical CRITICAL Values for the critical checks


SPECIAL NOTES :
Backends Default: warning --> 80% , critical --> 85% , of the number of connections
Accepts both numbers and percentages, and could be combined with an or
e.g 20 or 10% is allowed

Vacuum/analyze Default: warning --> 1 month. No default critical value.
Accepts : integers with the following units
day(s), week(s), month(s), year(s)

Autoanalyze/autovacuum Same as for vacuum and analyze

Relation sizes (tables,indexes and databases). Warning and/or critical value must be provided
Accepts integers with the following units
KB,MB,GB,TB,PB

Bloat Default : warning --> 1GB. No default critical value provided
Accepts integers with the following units
KB,MB,GB,TB,PB

Locks Default : warning --> 1min . No default critical value provided
Accepts integers with the smallest units being minute(s)
** accept others higher than that, but they don't make any
technical sence and might not be useful for the purpose of
lock monitoring.

Replica_lag Default : warning --> 5. Critical --> 10 . (16MB worth of WAL files)
Accepts only integers. Units not allowed

WALS No defaults. Warning and/or critical value must be provided
Only integer accepted




SAMPLE USAGE :
Help pg_monitor --help

Checking the number of backends
pg_monitor --check=backends --user=monitor --password=secrete --dbname=mydbName --host=hostname

*** Depending on your postgres hba settings, password may not be necessary.

pg_monitor --check=backends --user=monitor --password=secrete --dbname=mydbName --host=hostname --warning='90 or 85%' --critical='95%'



Checking for replica lag
pg_monitor --check=replica_lag --user=monitor --password=secrete --dbname=mydbName --host=master,replica1,replica2,replicaN --port=master,replica1,replica2,replicaN