27 Commits

Author SHA1 Message Date
3ade30c04a Add script to shot tempalte coverage
* Add a script to compare the defined metrics with those used in the
  Zabbix template.

* Add a Makefile target to run the above script.
2025-10-09 02:21:17 -04:00
ab039dc412 Improve test coverage
* Refactor some functions to improve unit test coverage.

* Use the "coverage" module if available along with unit tests.
2025-10-08 23:53:45 -04:00
75c5d76047 Add a metric for tracking the number of granted locks 2025-09-23 01:14:50 -04:00
43cd162313 Refactor to make pylint happy 2025-09-23 01:12:49 -04:00
29bfd07dad Run code through black 2025-07-14 01:58:08 -04:00
60589c2058 Update undiscovered item behavior 2025-07-14 01:39:57 -04:00
ea3aca3455 Filter out initial logical replication sync workers
* Slots are crated for each table during the initial sync, which only
  live for the duration of the copy for that table.

* The initial sync backend workers' names are also based on the table
  being copied.

* Filter out the above in Zabbix discovery based on application_name and
  slot_name.
2025-07-07 13:15:03 -04:00
cc71547f5f Include application_name in replication discovery data 2025-07-07 13:07:11 -04:00
107d5056d6 Merge branch 'main' into develop 2025-07-06 03:37:37 -04:00
e930178e9c Merge branch 'rel/1.0.4' 2025-07-06 03:35:39 -04:00
5afc940df8 Bump release version to 1.0.4 2025-07-06 03:35:00 -04:00
3961aa3448 Add Gentoo ebuild for 1.0.4-rc1 2025-07-06 03:32:27 -04:00
8fd57032e7 Update Makefile to support rc versions, bump version
* Support packaging RC versions for deb and rpm packages

* Bump version to 1.0.4-rc1
2025-07-06 03:28:36 -04:00
5ede7dea07 Fix package-all make target
* Fix the package-all target in the Makefile

* Remove storage of raw sequence stats
2025-07-05 12:28:36 -04:00
5ea007c3f6 Merge branch 'dev/sequences' into develop 2025-07-05 01:18:01 -04:00
7cb0f7ad40 Fix typo in activity query 2025-07-05 01:17:35 -04:00
83fa12ec54 Correct types for slot LSN lag metrics 2025-07-04 02:46:25 -04:00
98b74d9aed Add sequence metrics to Zabbix template 2025-07-04 02:45:43 -04:00
45953848e2 Remove some type casts from working around Decimal types 2025-07-03 10:16:20 -04:00
6116f4f885 Fix missing json default 2025-07-03 02:08:45 -04:00
24d1214855 Teach json how to serialize decimals 2025-07-03 01:47:06 -04:00
3c39d8aa97 Remove CIDR prefix from replication id 2025-07-03 01:13:58 -04:00
ebb084aa9d Add initial query for sequence usage 2025-07-01 02:29:56 -04:00
86d5e8917b Merge branch 'dev/io_stats' into develop 2025-07-01 01:30:21 -04:00
be84e2b89a Report some I/O metrics in bytes 2025-06-30 02:00:06 -04:00
fd0a9e1230 Add backend I/O discoovery items to template 2025-06-29 02:08:19 -04:00
b934a30124 Add pg_stat_io query
* Add a query for pg_stat_io metrics grouped by backend type

* Switch some unsupported values from 0 to NULL to better reflect their
  nonexistence
2025-06-28 00:30:03 -04:00
11 changed files with 2021 additions and 469 deletions

1
.gitignore vendored
View File

@@ -8,3 +8,4 @@ missing
__pycache__
venv
build
.coverage

74
GENTOO/pgmon-1.0.4.ebuild Normal file
View File

@@ -0,0 +1,74 @@
# Copyright 2024 Gentoo Authors
# Distributed under the terms of the GNU General Public License v2
EAPI=8
PYTHON_COMPAT=( python3_{6..13} )
inherit python-r1 systemd
DESCRIPTION="PostgreSQL monitoring bridge"
HOMEPAGE="None"
LICENSE="BSD"
SLOT="0"
KEYWORDS="amd64"
SRC_URI="https://code2.shh-dot-com.org/james/${PN}/releases/download/v${PV}/${P}.tar.bz2"
IUSE="-systemd"
DEPEND="
${PYTHON_DEPS}
dev-python/psycopg:2
dev-python/pyyaml
dev-python/requests
app-admin/logrotate
"
RDEPEND="${DEPEND}"
BDEPEND=""
#RESTRICT="fetch"
#S="${WORKDIR}/${PN}"
#pkg_nofetch() {
# einfo "Please download"
# einfo " - ${P}.tar.bz2"
# einfo "from ${HOMEPAGE} and place it in your DISTDIR directory."
# einfo "The file should be owned by portage:portage."
#}
src_compile() {
true
}
src_install() {
# Install init script
if ! use systemd ; then
newinitd "openrc/pgmon.initd" pgmon
newconfd "openrc/pgmon.confd" pgmon
fi
# Install systemd unit
if use systemd ; then
systemd_dounit "systemd/pgmon.service"
fi
# Install script
exeinto /usr/bin
newexe "src/pgmon.py" pgmon
# Install default config
diropts -o root -g root -m 0755
insinto /etc/pgmon
doins "sample-config/pgmon.yml"
doins "sample-config/pgmon-metrics.yml"
# Install logrotate config
insinto /etc/logrotate.d
newins "logrotate/pgmon.logrotate" pgmon
# Install man page
doman manpages/pgmon.1
}

View File

@@ -3,7 +3,28 @@ PACKAGE_NAME := pgmon
SCRIPT := src/$(PACKAGE_NAME).py
VERSION := $(shell grep -m 1 '^VERSION = ' "$(SCRIPT)" | sed -ne 's/.*"\(.*\)".*/\1/p')
# Figure out the version components
# Note: The release is for RPM packages, where prerelease releases are written as 0.<release>
FULL_VERSION := $(shell grep -m 1 '^VERSION = ' "$(SCRIPT)" | sed -ne 's/.*"\(.*\)".*/\1/p')
VERSION := $(shell echo $(FULL_VERSION) | sed -n 's/\(.*\)\(-rc.*\|$$\)/\1/p')
RELEASE := $(shell echo $(FULL_VERSION) | sed -n 's/.*-rc\([0-9]\+\)$$/\1/p')
ifeq ($(RELEASE),)
RPM_RELEASE := 1
RPM_VERSION := $(VERSION)-$(RPM_RELEASE)
DEB_VERSION := $(VERSION)
else
RPM_RELEASE := 0.$(RELEASE)
RPM_VERSION := $(VERSION)-$(RPM_RELEASE)
DEB_VERSION := $(VERSION)~rc$(RELEASE)
endif
# Python stuff
PYTHON ?= python3
PYLINT ?= pylint
BLACK ?= black
HAVE_COVERAGE := $(shell $(PYTHON) -c 'import coverage' 2>/dev/null && echo yes)
# Where packages are built
BUILD_DIR := build
@@ -23,18 +44,22 @@ SUPPORTED := ubuntu-20.04 \
# These targets are the main ones to use for most things.
##
.PHONY: all clean tgz test query-tests install-common install-openrc install-systemd
.PHONY: all clean tgz lint format test query-tests install-common install-openrc install-systemd
all: package-all
version:
@echo "full version=$(FULL_VERSION) version=$(VERSION) rel=$(RELEASE) rpm=$(RPM_VERSION) deb=$(DEB_VERSION)"
# Build all packages
.PHONY: package-all
all: $(foreach distro_release, $(SUPPORTED), package-$(distro_release))
package-all: $(foreach distro_release, $(SUPPORTED), package-$(distro_release))
# Gentoo package (tar.gz) creation
.PHONY: package-gentoo
package-gentoo:
mkdir -p $(BUILD_DIR)/gentoo
tar --transform "s,^,$(PACKAGE_NAME)-$(VERSION)/," -acjf $(BUILD_DIR)/gentoo/$(PACKAGE_NAME)-$(VERSION).tar.bz2 --exclude .gitignore $(shell git ls-tree --full-tree --name-only -r HEAD)
tar --transform "s,^,$(PACKAGE_NAME)-$(FULL_VERSION)/," -acjf $(BUILD_DIR)/gentoo/$(PACKAGE_NAME)-$(FULL_VERSION).tar.bz2 --exclude .gitignore $(shell git ls-tree --full-tree --name-only -r HEAD)
# Create a deb package
@@ -55,20 +80,41 @@ tgz:
rm -rf $(BUILD_DIR)/tgz/root
mkdir -p $(BUILD_DIR)/tgz/root
$(MAKE) install-openrc DESTDIR=$(BUILD_DIR)/tgz/root
tar -cz -f $(BUILD_DIR)/tgz/$(PACKAGE_NAME)-$(VERSION).tgz -C $(BUILD_DIR)/tgz/root .
tar -cz -f $(BUILD_DIR)/tgz/$(PACKAGE_NAME)-$(FULL_VERSION).tgz -C $(BUILD_DIR)/tgz/root .
# Clean up the build directory
clean:
rm -rf $(BUILD_DIR)
# Check for lint
lint:
$(PYLINT) src/pgmon.py
$(PYLINT) src/test_pgmon.py
$(BLACK) --check --diff src/pgmon.py
$(BLACK) --check --diff src/test_pgmon.py
# Format the code using black
format:
$(BLACK) src/pgmon.py
$(BLACK) src/test_pgmon.py
# Run unit tests for the script
test:
cd src ; python3 -m unittest
ifeq ($(HAVE_COVERAGE),yes)
cd src ; $(PYTHON) -m coverage run -m unittest && python3 -m coverage report -m
else
cd src ; $(PYTHON) -m unittest
endif
# Run query tests
query-tests:
cd tests ; ./run-tests.sh
# Compare the sample metrics with the Zabbix template
template-coverage:
$(PYTHON) zabbix_templates/coverage.py sample-config/pgmon-metrics.yml zabbix_templates/pgmon_templates.yaml
# Install the script at the specified base directory (common components)
install-common:
# Set up directories
@@ -129,28 +175,28 @@ debian-%-install-test:
docker run --rm \
-v ./$(BUILD_DIR):/output \
debian:$* \
bash -c 'apt-get update && apt-get install -y /output/$(PACKAGE_NAME)-$(VERSION)-debian-$*.deb'
bash -c 'apt-get update && apt-get install -y /output/$(PACKAGE_NAME)-$(DEB_VERSION)-debian-$*.deb'
# Run a RedHat install test
rockylinux-%-install-test:
docker run --rm \
-v ./$(BUILD_DIR):/output \
rockylinux:$* \
bash -c 'dnf makecache && dnf install -y /output/$(PACKAGE_NAME)-$(VERSION)-1.el$*.noarch.rpm'
bash -c 'dnf makecache && dnf install -y /output/$(PACKAGE_NAME)-$(RPM_VERSION).el$*.noarch.rpm'
# Run an Ubuntu install test
ubuntu-%-install-test:
docker run --rm \
-v ./$(BUILD_DIR):/output \
ubuntu:$* \
bash -c 'apt-get update && apt-get install -y /output/$(PACKAGE_NAME)-$(VERSION)-ubuntu-$*.deb'
bash -c 'apt-get update && apt-get install -y /output/$(PACKAGE_NAME)-$(DEB_VERSION)-ubuntu-$*.deb'
# Run an OracleLinux install test (this is for EL7 since CentOS7 images no longer exist)
oraclelinux-%-install-test:
docker run --rm \
-v ./$(BUILD_DIR):/output \
oraclelinux:7 \
bash -c 'yum makecache && yum install -y /output/$(PACKAGE_NAME)-$(VERSION)-1.el7.noarch.rpm'
bash -c 'yum makecache && yum install -y /output/$(PACKAGE_NAME)-$(RPM_VERSION).el7.noarch.rpm'
# Run a Gentoo install test
gentoo-install-test:
@@ -192,28 +238,28 @@ package-image-%:
actually-package-debian-%:
$(MAKE) install-systemd DESTDIR=/output/debian-$*
cp -r --preserve=mode DEBIAN /output/debian-$*/
dpkg-deb -Zgzip --build /output/debian-$* "/output/$(PACKAGE_NAME)-$(VERSION)-debian-$*.deb"
dpkg-deb -Zgzip --build /output/debian-$* "/output/$(PACKAGE_NAME)-$(DEB_VERSION)-debian-$*.deb"
# RedHat package creation
actually-package-rockylinux-%:
mkdir -p /output/rockylinux-$*/{BUILD,RPMS,SOURCES,SPECS,SRPMS}
sed -e "s/@@VERSION@@/$(VERSION)/g" RPM/$(PACKAGE_NAME).spec > /output/rockylinux-$*/SPECS/$(PACKAGE_NAME).spec
sed -e "s/@@VERSION@@/$(VERSION)/g" -e "s/@@RELEASE@@/$(RPM_RELEASE)/g" RPM/$(PACKAGE_NAME).spec > /output/rockylinux-$*/SPECS/$(PACKAGE_NAME).spec
rpmbuild --define '_topdir /output/rockylinux-$*' \
--define 'version $(VERSION)' \
--define 'version $(RPM_VERSION)' \
-bb /output/rockylinux-$*/SPECS/$(PACKAGE_NAME).spec
cp /output/rockylinux-$*/RPMS/noarch/$(PACKAGE_NAME)-$(VERSION)-1.el$*.noarch.rpm /output/
cp /output/rockylinux-$*/RPMS/noarch/$(PACKAGE_NAME)-$(RPM_VERSION).el$*.noarch.rpm /output/
# Ubuntu package creation
actually-package-ubuntu-%:
$(MAKE) install-systemd DESTDIR=/output/ubuntu-$*
cp -r --preserve=mode DEBIAN /output/ubuntu-$*/
dpkg-deb -Zgzip --build /output/ubuntu-$* "/output/$(PACKAGE_NAME)-$(VERSION)-ubuntu-$*.deb"
dpkg-deb -Zgzip --build /output/ubuntu-$* "/output/$(PACKAGE_NAME)-$(DEB_VERSION)-ubuntu-$*.deb"
# OracleLinux package creation
actually-package-oraclelinux-%:
mkdir -p /output/oraclelinux-$*/{BUILD,RPMS,SOURCES,SPECS,SRPMS}
sed -e "s/@@VERSION@@/$(VERSION)/g" RPM/$(PACKAGE_NAME)-el7.spec > /output/oraclelinux-$*/SPECS/$(PACKAGE_NAME).spec
sed -e "s/@@VERSION@@/$(VERSION)/g" -e "s/@@RELEASE@@/$(RPM_RELEASE)/g" RPM/$(PACKAGE_NAME)-el7.spec > /output/oraclelinux-$*/SPECS/$(PACKAGE_NAME).spec
rpmbuild --define '_topdir /output/oraclelinux-$*' \
--define 'version $(VERSION)' \
--define 'version $(RPM_VERSION)' \
-bb /output/oraclelinux-$*/SPECS/$(PACKAGE_NAME).spec
cp /output/oraclelinux-$*/RPMS/noarch/$(PACKAGE_NAME)-$(VERSION)-1.el$*.noarch.rpm /output/
cp /output/oraclelinux-$*/RPMS/noarch/$(PACKAGE_NAME)-$(RPM_VERSION).el$*.noarch.rpm /output/

View File

@@ -1,6 +1,6 @@
Name: pgmon
Version: @@VERSION@@
Release: 1%{?dist}
Release: @@RELEASE@@%{?dist}
Summary: A bridge to sit between monitoring tools and PostgreSQL
License: MIT

View File

@@ -1,6 +1,6 @@
Name: pgmon
Version: @@VERSION@@
Release: 1%{?dist}
Release: @@RELEASE@@%{?dist}
Summary: A bridge to sit between monitoring tools and PostgreSQL
License: MIT

4
pylintrc Normal file
View File

@@ -0,0 +1,4 @@
[MASTER]
py-version=3.5
disable=fixme

View File

@@ -8,14 +8,22 @@ metrics:
0: >
SELECT datname AS dbname
FROM pg_database
# Note: If the user lacks sufficient privileges, these fields will be NULL.
# The WHERE clause is intended to prevent Zabbix from discovering a
# connection it cannot monitor. Ideally this would generate an error
# instead.
discover_rep:
type: set
query:
0: >
SELECT client_addr || '_' || regexp_replace(application_name, '[ ,]', '_', 'g') AS repid,
SELECT host(client_addr) || '_' || regexp_replace(application_name, '[ ,]', '_', 'g') AS repid,
application_name,
client_addr,
state
FROM pg_stat_replication
WHERE state IS NOT NULL
discover_slots:
type: set
query:
@@ -36,6 +44,7 @@ metrics:
active
FROM pg_replication_slots
##
# cluster-wide metrics
##
@@ -49,7 +58,7 @@ metrics:
query:
0: >
SELECT max(age(datfrozenxid)) AS xid_age,
0 AS mxid_age
NULL AS mxid_age
FROM pg_database
90600: >
SELECT max(age(datfrozenxid)) AS xid_age,
@@ -85,6 +94,28 @@ metrics:
FROM pg_stat_bgwriter bg
CROSS JOIN pg_stat_checkpointer cp
io_per_backend:
type: set
query:
160000: >
SELECT backend_type,
COALESCE(SUM(reads * op_bytes), 0) AS reads,
COALESCE(SUM(read_time), 0) AS read_time,
COALESCE(SUM(writes * op_bytes), 0) AS writes,
COALESCE(SUM(write_time), 0) AS write_time,
COALESCE(SUM(writebacks * op_bytes), 0) AS writebacks,
COALESCE(SUM(writeback_time), 0) AS writeback_time,
COALESCE(SUM(extends * op_bytes), 0) AS extends,
COALESCE(SUM(extend_time), 0) AS extend_time,
COALESCE(SUM(op_bytes), 0) AS op_bytes,
COALESCE(SUM(hits), 0) AS hits,
COALESCE(SUM(evictions), 0) AS evictions,
COALESCE(SUM(reuses), 0) AS reuses,
COALESCE(SUM(fsyncs), 0) AS fsyncs,
COALESCE(SUM(fsync_time), 0) AS fsync_time
FROM pg_stat_io
GROUP BY backend_type
##
# Per-database metrics
@@ -107,17 +138,17 @@ metrics:
temp_files,
temp_bytes,
deadlocks,
0 AS checksum_failures,
NULL AS checksum_failures,
blk_read_time,
blk_write_time,
0 AS session_time,
0 AS active_time,
0 AS idle_in_transaction_time,
0 AS sessions,
0 AS sessions_abandoned,
0 AS sessions_fatal,
0 AS sessions_killed,
extract('epoch' from stats_reset)::float AS stats_reset
NULL AS session_time,
NULL AS active_time,
NULL AS idle_in_transaction_time,
NULL AS sessions,
NULL AS sessions_abandoned,
NULL AS sessions_fatal,
NULL AS sessions_killed,
extract('epoch' from stats_reset) AS stats_reset
FROM pg_stat_database WHERE datname = %(dbname)s
140000: >
SELECT numbackends,
@@ -134,8 +165,7 @@ metrics:
temp_files,
temp_bytes,
deadlocks,
COALESCE(checksum_failures,
0) AS checksum_failures,
COALESCE(checksum_failures, 0) AS checksum_failures,
blk_read_time,
blk_write_time,
session_time,
@@ -145,7 +175,7 @@ metrics:
sessions_abandoned,
sessions_fatal,
sessions_killed,
extract('epoch' from stats_reset)::float AS stats_reset
extract('epoch' from stats_reset) AS stats_reset
FROM pg_stat_database WHERE datname = %(dbname)s
test_args:
dbname: postgres
@@ -168,13 +198,52 @@ metrics:
0: >
SELECT state,
count(*) AS backend_count,
COALESCE(EXTRACT(EPOCH FROM max(now() - state_change))::float, 0) AS max_state_time
COALESCE(EXTRACT(EPOCH FROM max(now() - state_change)), 0) AS max_state_time
FROM pg_stat_activity
WHERE datname = %(dbname)s
GROUP BY state
test_args:
dbname: postgres
sequence_usage:
type: value
query:
# 9.2 lacks lateral joins, the pg_sequence_last_value function, and the pg_sequences view
# 0: >
# SELECT COALESCE(MAX(pg_sequence_last_value(c.oid)::float / (pg_sequence_parameters(oid)).maximum_value), 0) AS max_usage
# FROM pg_class c
# WHERE c.relkind = 'S'
# 9.3 - 9.6 lacks the pg_sequence_last_value function, and pg_sequences view
# 90300: >
# SELECT COALESCE(MAX(pg_sequence_last_value(c.oid)::float / s.maximum_value), 0) AS max_usage
# FROM pg_class c
# CROSS JOIN LATERAL pg_sequence_parameters(c.oid) AS s
# WHERE c.relkind = 'S'
100000: SELECT COALESCE(MAX(last_value::float / max_value), 0) AS max_usage FROM pg_sequences;
test_args:
dbname: postgres
sequence_visibility:
type: row
query:
100000: >
SELECT COUNT(*) FILTER (WHERE has_sequence_privilege(c.oid, 'SELECT,USAGE')) AS visible_sequences,
COUNT(*) AS total_sequences
FROM pg_class AS c
WHERE relkind = 'S'
locks:
type: row
query:
0:
SELECT COUNT(*) AS total,
SUM(CASE WHEN granted THEN 1 ELSE 0 END) AS granted
FROM pg_locks
90400: >
SELECT COUNT(*) AS total,
COUNT(*) FILTER (WHERE granted) AS granted
FROM pg_locks
##
# Per-replication metrics
@@ -184,7 +253,7 @@ metrics:
query:
90400: >
SELECT pid, usename,
EXTRACT(EPOCH FROM backend_start)::integer AS backend_start,
EXTRACT(EPOCH FROM backend_start) AS backend_start,
state,
pg_xlog_location_diff(pg_current_xlog_location(), sent_location) AS sent_lsn,
pg_xlog_location_diff(pg_current_xlog_location(), write_location) AS write_lsn,
@@ -195,20 +264,21 @@ metrics:
NULL AS replay_lag,
sync_state
FROM pg_stat_replication
WHERE client_addr || '_' || regexp_replace(application_name, '[ ,]', '_', 'g') = %(repid)s
WHERE host(client_addr) || '_' || regexp_replace(application_name, '[ ,]', '_', 'g') = %(repid)s
100000: >
SELECT pid, usename,
EXTRACT(EPOCH FROM backend_start)::integer AS backend_start,
EXTRACT(EPOCH FROM backend_start) AS backend_start,
state,
pg_wal_lsn_diff(pg_current_wal_lsn(), sent_lsn) AS sent_lsn,
pg_wal_lsn_diff(pg_current_wal_lsn(), write_lsn) AS write_lsn,
pg_wal_lsn_diff(pg_current_wal_lsn(), flush_lsn) AS flush_lsn,
pg_wal_lsn_diff(pg_current_wal_lsn(), replay_lsn) AS replay_lsn,
COALESCE(EXTRACT(EPOCH FROM write_lag), 0)::integer AS write_lag,
COALESCE(EXTRACT(EPOCH FROM flush_lag), 0)::integer AS flush_lag,
COALESCE(EXTRACT(EPOCH FROM replay_lag), 0)::integer AS replay_lag,
COALESCE(EXTRACT(EPOCH FROM write_lag), 0) AS write_lag,
COALESCE(EXTRACT(EPOCH FROM flush_lag), 0) AS flush_lag,
COALESCE(EXTRACT(EPOCH FROM replay_lag), 0) AS replay_lag,
sync_state
FROM pg_stat_replication WHERE client_addr || '_' || regexp_replace(application_name, '[ ,]', '_', 'g') = %(repid)s
FROM pg_stat_replication
WHERE host(client_addr) || '_' || regexp_replace(application_name, '[ ,]', '_', 'g') = %(repid)s
test_args:
repid: 127.0.0.1_test_rep
@@ -240,6 +310,7 @@ metrics:
test_args:
slot: test_slot
##
# Debugging
##

View File

@@ -1,103 +1,147 @@
#!/usr/bin/env python3
"""
pgmon is a monitoring intermediary that sits between a PostgreSQL cluster and a monitoring systen
that is capable of parsing JSON responses over an HTTP connection.
"""
# pylint: disable=too-few-public-methods
import yaml
import json
import time
import os
import sys
import signal
import argparse
import logging
import re
from decimal import Decimal
from urllib.parse import urlparse, parse_qs
from contextlib import contextmanager
from datetime import datetime, timedelta
from http.server import BaseHTTPRequestHandler
from http.server import ThreadingHTTPServer
from threading import Lock
import yaml
import psycopg2
from psycopg2.extras import RealDictCursor
from psycopg2.pool import ThreadedConnectionPool
from contextlib import contextmanager
import signal
from threading import Thread, Lock, Semaphore
from http.server import BaseHTTPRequestHandler, HTTPServer
from http.server import ThreadingHTTPServer
from urllib.parse import urlparse, parse_qs
import requests
import re
VERSION = "1.0.3"
# Configuration
config = {}
VERSION = "1.0.4"
# Dictionary of current PostgreSQL connection pools
connections_lock = Lock()
connections = {}
# Dictionary of unhappy databases. Keys are database names, value is the time
# the database was determined to be unhappy plus the cooldown setting. So,
# basically it's the time when we should try to connect to the database again.
unhappy_cooldown = {}
class Context:
"""
The global context for connections, config, version, nad IPC
"""
# Version information
cluster_version = None
cluster_version_next_check = None
cluster_version_lock = Lock()
# Configuration
config = {}
# PostgreSQL latest version information
latest_version = None
latest_version_next_check = None
latest_version_lock = Lock()
release_supported = None
# Dictionary of current PostgreSQL connection pools
connections_lock = Lock()
connections = {}
# Running state (used to gracefully shut down)
running = True
# Dictionary of unhappy databases. Keys are database names, value is the time
# the database was determined to be unhappy plus the cooldown setting. So,
# basically it's the time when we should try to connect to the database again.
unhappy_cooldown = {}
# The http server object
httpd = None
# Version information
cluster_version = None
cluster_version_next_check = None
cluster_version_lock = Lock()
# Where the config file lives
config_file = None
# PostgreSQL latest version information
latest_version = None
latest_version_next_check = None
latest_version_lock = Lock()
release_supported = None
# Configure logging
log = logging.getLogger(__name__)
formatter = logging.Formatter(
"%(asctime)s - %(levelname)s - %(filename)s: %(funcName)s() line %(lineno)d: %(message)s"
)
console_log_handler = logging.StreamHandler()
console_log_handler.setFormatter(formatter)
log.addHandler(console_log_handler)
# Running state (used to gracefully shut down)
running = True
# The http server object
httpd = None
# Where the config file lives
config_file = None
# Configure logging
log = logging.getLogger(__name__)
@classmethod
def init_logging(cls):
"""
Actually initialize the logging framework. Since we don't ever instantiate the Context
class, this provides a way to make a few modifications to the log handler.
"""
formatter = logging.Formatter(
"%(asctime)s - %(levelname)s - %(filename)s: "
"%(funcName)s() line %(lineno)d: %(message)s"
)
console_log_handler = logging.StreamHandler()
console_log_handler.setFormatter(formatter)
cls.log.addHandler(console_log_handler)
# Error types
class ConfigError(Exception):
pass
"""
Error type for all config related errors.
"""
class DisconnectedError(Exception):
pass
"""
Error indicating a previously active connection to the database has been disconnected.
"""
class UnhappyDBError(Exception):
pass
"""
Error indicating that a database the code has been asked to connect to is on the unhappy list.
"""
class UnknownMetricError(Exception):
pass
"""
Error indicating that an undefined metric was requested.
"""
class MetricVersionError(Exception):
pass
"""
Error indicating that there is no suitable query for a metric that was requested for the
version of PostgreSQL being monitored.
"""
class LatestVersionCheckError(Exception):
pass
"""
Error indicating that there was a problem retrieving or parsing the latest version information.
"""
class InvalidDataError(Exception):
"""
Error indicating query results were somehow invalid
"""
# Default config settings
default_config = {
DEFAULT_CONFIG = {
# The address the agent binds to
"address": "127.0.0.1",
# The port the agent listens on for requests
@@ -175,6 +219,60 @@ def update_deep(d1, d2):
return d1
def validate_metric(path, name, metric):
"""
Validate a metric definition from a given file. If any query definitions come from external
files, the metric dict will be updated with the actual query.
Params:
path: path to the file which contains this definition
name: name of the metric
metric: the dictionary containing the metric definition
"""
# Validate return types
try:
if metric["type"] not in ["value", "row", "column", "set"]:
raise ConfigError(
"Invalid return type: {} for metric {} in {}".format(
metric["type"], name, path
)
)
except KeyError as e:
raise ConfigError(
"No type specified for metric {} in {}".format(name, path)
) from e
# Ensure queries exist
query_dict = metric.get("query", {})
if not isinstance(query_dict, dict):
raise ConfigError(
"Query definition should be a dictionary, got: {} for metric {} in {}".format(
query_dict, name, path
)
)
if len(query_dict) == 0:
raise ConfigError("Missing queries for metric {} in {}".format(name, path))
# Read external sql files and validate version keys
config_base = os.path.dirname(path)
for vers, query in metric["query"].items():
try:
int(vers)
except Exception as e:
raise ConfigError(
"Invalid version: {} for metric {} in {}".format(vers, name, path)
) from e
# Read in the external query and update the definition in the metricdictionary
if query.startswith("file:"):
query_path = query[5:]
if not query_path.startswith("/"):
query_path = os.path.join(config_base, query_path)
with open(query_path, "r", encoding="utf-8") as f:
metric["query"][vers] = f.read()
def read_config(path, included=False):
"""
Read a config file.
@@ -183,61 +281,21 @@ def read_config(path, included=False):
path: path to the file to read
included: is this file included by another file?
"""
# Read config file
log.info("Reading log file: {}".format(path))
with open(path, "r") as f:
Context.log.info("Reading log file: %s", path)
with open(path, "r", encoding="utf-8") as f:
try:
cfg = yaml.safe_load(f)
except yaml.parser.ParserError as e:
raise ConfigError("Inavlid config file: {}: {}".format(path, e))
# Since we use it a few places, get the base directory from the config
config_base = os.path.dirname(path)
raise ConfigError("Inavlid config file: {}: {}".format(path, e)) from e
# Read any external queries and validate metric definitions
for name, metric in cfg.get("metrics", {}).items():
# Validate return types
try:
if metric["type"] not in ["value", "row", "column", "set"]:
raise ConfigError(
"Invalid return type: {} for metric {} in {}".format(
metric["type"], name, path
)
)
except KeyError:
raise ConfigError(
"No type specified for metric {} in {}".format(name, path)
)
# Ensure queries exist
query_dict = metric.get("query", {})
if type(query_dict) is not dict:
raise ConfigError(
"Query definition should be a dictionary, got: {} for metric {} in {}".format(
query_dict, name, path
)
)
if len(query_dict) == 0:
raise ConfigError("Missing queries for metric {} in {}".format(name, path))
# Read external sql files and validate version keys
for vers, query in metric["query"].items():
try:
int(vers)
except:
raise ConfigError(
"Invalid version: {} for metric {} in {}".format(vers, name, path)
)
if query.startswith("file:"):
query_path = query[5:]
if not query_path.startswith("/"):
query_path = os.path.join(config_base, query_path)
with open(query_path, "r") as f:
metric["query"][vers] = f.read()
validate_metric(path, name, metric)
# Read any included config files
config_base = os.path.dirname(path)
for inc in cfg.get("include", []):
# Prefix relative paths with the directory from the current config
if not inc.startswith("/"):
@@ -248,34 +306,37 @@ def read_config(path, included=False):
# config
if included:
return cfg
else:
new_config = {}
update_deep(new_config, default_config)
update_deep(new_config, cfg)
# Minor sanity checks
if len(new_config["metrics"]) == 0:
log.error("No metrics are defined")
raise ConfigError("No metrics defined")
new_config = {}
update_deep(new_config, DEFAULT_CONFIG)
update_deep(new_config, cfg)
# Validate the new log level before changing the config
if new_config["log_level"].upper() not in [
"DEBUG",
"INFO",
"WARNING",
"ERROR",
"CRITICAL",
]:
raise ConfigError("Invalid log level: {}".format(new_config["log_level"]))
# Minor sanity checks
if len(new_config["metrics"]) == 0:
Context.log.error("No metrics are defined")
raise ConfigError("No metrics defined")
global config
config = new_config
# Validate the new log level before changing the config
if new_config["log_level"].upper() not in [
"DEBUG",
"INFO",
"WARNING",
"ERROR",
"CRITICAL",
]:
raise ConfigError("Invalid log level: {}".format(new_config["log_level"]))
# Apply changes to log level
log.setLevel(logging.getLevelName(config["log_level"].upper()))
Context.config = new_config
# Apply changes to log level
Context.log.setLevel(logging.getLevelName(Context.config["log_level"].upper()))
# Return the config (mostly to make pylint happy, but also in case I opt to remove the side
# effect and make this more functional.
return Context.config
def signal_handler(sig, frame):
def signal_handler(sig, frame): # pylint: disable=unused-argument
"""
Function for handling signals
@@ -286,19 +347,22 @@ def signal_handler(sig, frame):
# Signal everything to shut down
if sig in [signal.SIGINT, signal.SIGTERM, signal.SIGQUIT]:
log.info("Shutting down ...")
global running
running = False
if httpd is not None:
httpd.socket.close()
Context.log.info("Shutting down ...")
Context.running = False
if Context.httpd is not None:
Context.httpd.socket.close()
# Signal a reload
if sig == signal.SIGHUP:
log.warning("Received config reload signal")
read_config(config_file)
Context.log.warning("Received config reload signal")
read_config(Context.config_file)
class ConnectionPool(ThreadedConnectionPool):
"""
Threaded connection pool that has a context manager.
"""
def __init__(self, dbname, minconn, maxconn, *args, **kwargs):
# Make sure dbname isn't different in the kwargs
kwargs["dbname"] = dbname
@@ -307,7 +371,14 @@ class ConnectionPool(ThreadedConnectionPool):
self.name = dbname
@contextmanager
def connection(self, timeout=None):
def connection(self, timeout):
"""
Connection context manager for our connection pool. This will attempt to retrieve a
connection until the timeout is reached.
Params:
timeout: how long to keep trying to get a connection bedore giving up
"""
conn = None
timeout_time = datetime.now() + timedelta(timeout)
# We will continue to try to get a connection slot until we time out
@@ -331,34 +402,37 @@ class ConnectionPool(ThreadedConnectionPool):
def get_pool(dbname):
"""
Get a database connection pool.
Params:
dbname: the name of the database for which a connection pool should be returned.
"""
# Check if the db is unhappy and wants to be left alone
if dbname in unhappy_cooldown:
if unhappy_cooldown[dbname] > datetime.now():
if dbname in Context.unhappy_cooldown:
if Context.unhappy_cooldown[dbname] > datetime.now():
raise UnhappyDBError()
# Create a connection pool if it doesn't already exist
if dbname not in connections:
with connections_lock:
if dbname not in Context.connections:
with Context.connections_lock:
# Make sure nobody created the pool while we were waiting on the
# lock
if dbname not in connections:
log.info("Creating connection pool for: {}".format(dbname))
if dbname not in Context.connections:
Context.log.info("Creating connection pool for: %s", dbname)
# Actually create the connection pool
connections[dbname] = ConnectionPool(
Context.connections[dbname] = ConnectionPool(
dbname,
int(config["min_pool_size"]),
int(config["max_pool_size"]),
int(Context.config["min_pool_size"]),
int(Context.config["max_pool_size"]),
application_name="pgmon",
host=config["dbhost"],
port=config["dbport"],
user=config["dbuser"],
connect_timeout=int(config["connect_timeout"]),
sslmode=config["ssl_mode"],
host=Context.config["dbhost"],
port=Context.config["dbport"],
user=Context.config["dbuser"],
connect_timeout=int(Context.config["connect_timeout"]),
sslmode=Context.config["ssl_mode"],
)
# Clear the unhappy indicator if present
unhappy_cooldown.pop(dbname, None)
return connections[dbname]
Context.unhappy_cooldown.pop(dbname, None)
return Context.connections[dbname]
def handle_connect_failure(pool):
@@ -366,8 +440,8 @@ def handle_connect_failure(pool):
Mark the database as being unhappy so we can leave it alone for a while
"""
dbname = pool.name
unhappy_cooldown[dbname] = datetime.now() + timedelta(
seconds=int(config["reconnect_cooldown"])
Context.unhappy_cooldown[dbname] = datetime.now() + timedelta(
seconds=int(Context.config["reconnect_cooldown"])
)
@@ -391,38 +465,74 @@ def get_query(metric, version):
raise MetricVersionError("Missing metric query for PostgreSQL {}".format(version))
def json_encode_special(obj):
"""
Encoder function to handle types the standard JSON package doesn't know what
to do with
"""
if isinstance(obj, Decimal):
return float(obj)
raise TypeError("Cannot serialize object of {}".format(type(obj)))
def json_encode_result(return_type, res):
"""
Return a json string encoding of the results of a query.
params:
return_type: the expected structure to return. One of:
value, row, column, set
res: the query results
returns: a json string form of the results
raises:
ConfigError: when an invalid return_type is given
InvalidDataError: when the query results don't match the return type
"""
try:
if return_type == "value":
if len(res) == 0:
return ""
return str(list(res[0].values())[0])
if return_type == "row":
return json.dumps(
res[0] if len(res) > 0 else {}, default=json_encode_special
)
if return_type == "column":
return json.dumps(
[list(r.values())[0] for r in res], default=json_encode_special
)
if return_type == "set":
return json.dumps(res, default=json_encode_special)
except IndexError as e:
raise InvalidDataError(e) from e
# If we got to this point, the return type is invalid
raise ConfigError("Invalid query return type: {}".format(return_type))
def run_query_no_retry(pool, return_type, query, args):
"""
Run the query with no explicit retry code
"""
with pool.connection(float(config["connect_timeout"])) as conn:
with pool.connection(float(Context.config["connect_timeout"])) as conn:
try:
with conn.cursor(cursor_factory=RealDictCursor) as curs:
curs.execute(query, args)
res = curs.fetchall()
if return_type == "value":
if len(res) == 0:
return ""
return str(list(res[0].values())[0])
elif return_type == "row":
if len(res) == 0:
return "[]"
return json.dumps(res[0])
elif return_type == "column":
if len(res) == 0:
return "[]"
return json.dumps([list(r.values())[0] for r in res])
elif return_type == "set":
return json.dumps(res)
except:
return json_encode_result(return_type, res)
except Exception as e:
dbname = pool.name
if dbname in unhappy_cooldown:
raise UnhappyDBError()
elif conn.closed != 0:
raise DisconnectedError()
else:
raise
if dbname in Context.unhappy_cooldown:
raise UnhappyDBError() from e
if conn.closed != 0:
raise DisconnectedError() from e
raise
def run_query(pool, return_type, query, args):
@@ -443,7 +553,7 @@ def run_query(pool, return_type, query, args):
try:
return run_query_no_retry(pool, return_type, query, args)
except DisconnectedError:
log.warning("Stale PostgreSQL connection found ... trying again")
Context.log.warning("Stale PostgreSQL connection found ... trying again")
# This sleep is an annoying hack to give the pool workers time to
# actually mark the connection, otherwise it can be given back in the
# next connection() call
@@ -451,9 +561,9 @@ def run_query(pool, return_type, query, args):
time.sleep(1)
try:
return run_query_no_retry(pool, return_type, query, args)
except:
except Exception as e:
handle_connect_failure(pool)
raise UnhappyDBError()
raise UnhappyDBError() from e
def get_cluster_version():
@@ -461,40 +571,39 @@ def get_cluster_version():
Get the PostgreSQL version if we don't already know it, or if it's been
too long sice the last time it was checked.
"""
global cluster_version
global cluster_version_next_check
# If we don't know the version or it's past the recheck time, get the
# version from the database. Only one thread needs to do this, so they all
# try to grab the lock, and then make sure nobody else beat them to it.
if (
cluster_version is None
or cluster_version_next_check is None
or cluster_version_next_check < datetime.now()
Context.cluster_version is None
or Context.cluster_version_next_check is None
or Context.cluster_version_next_check < datetime.now()
):
with cluster_version_lock:
with Context.cluster_version_lock:
# Only check if nobody already got the version before us
if (
cluster_version is None
or cluster_version_next_check is None
or cluster_version_next_check < datetime.now()
Context.cluster_version is None
or Context.cluster_version_next_check is None
or Context.cluster_version_next_check < datetime.now()
):
log.info("Checking PostgreSQL cluster version")
pool = get_pool(config["dbname"])
cluster_version = int(
Context.log.info("Checking PostgreSQL cluster version")
pool = get_pool(Context.config["dbname"])
Context.cluster_version = int(
run_query(pool, "value", "SHOW server_version_num", None)
)
cluster_version_next_check = datetime.now() + timedelta(
seconds=int(config["version_check_period"])
Context.cluster_version_next_check = datetime.now() + timedelta(
seconds=int(Context.config["version_check_period"])
)
log.info("Got PostgreSQL cluster version: {}".format(cluster_version))
log.debug(
"Next PostgreSQL cluster version check will be after: {}".format(
cluster_version_next_check
)
Context.log.info(
"Got PostgreSQL cluster version: %s", Context.cluster_version
)
Context.log.debug(
"Next PostgreSQL cluster version check will be after: %s",
Context.cluster_version_next_check,
)
return cluster_version
return Context.cluster_version
def version_num_to_release(version_num):
@@ -507,8 +616,7 @@ def version_num_to_release(version_num):
"""
if version_num // 10000 < 10:
return version_num // 10000 + (version_num % 10000 // 100 / 10)
else:
return version_num // 10000
return version_num // 10000
def parse_version_rss(raw_rss, release):
@@ -516,7 +624,7 @@ def parse_version_rss(raw_rss, release):
Parse the raw RSS from the versions.rss feed to extract the latest version of
PostgreSQL that's availabe for the cluster being monitored.
This sets these global variables:
This sets these Context variables:
latest_version
release_supported
@@ -526,8 +634,6 @@ def parse_version_rss(raw_rss, release):
raw_rss: The raw rss text from versions.rss
release: The PostgreSQL release we care about (ex: 9.2, 14)
"""
global latest_version
global release_supported
# Regular expressions for parsing the RSS document
version_line = re.compile(
@@ -547,75 +653,75 @@ def parse_version_rss(raw_rss, release):
version = m.group(1)
parts = list(map(int, version.split(".")))
if parts[0] < 10:
latest_version = int(
Context.latest_version = int(
"{}{:02}{:02}".format(parts[0], parts[1], parts[2])
)
else:
latest_version = int("{}00{:02}".format(parts[0], parts[1]))
Context.latest_version = int("{}00{:02}".format(parts[0], parts[1]))
elif release_found:
# The next line after the version tells if the version is supported
if unsupported_line.match(line):
release_supported = False
Context.release_supported = False
else:
release_supported = True
Context.release_supported = True
break
# Make sure we actually found it
if not release_found:
raise LatestVersionCheckError("Current release ({}) not found".format(release))
log.info(
"Got latest PostgreSQL version: {} supported={}".format(
latest_version, release_supported
)
Context.log.info(
"Got latest PostgreSQL version: %s supported=%s",
Context.latest_version,
Context.release_supported,
)
log.debug(
"Next latest PostgreSQL version check will be after: {}".format(
latest_version_next_check
)
Context.log.debug(
"Next latest PostgreSQL version check will be after: %s",
Context.latest_version_next_check,
)
def get_latest_version():
"""
Get the latest supported version of the major PostgreSQL release running on the server being monitored.
Get the latest supported version of the major PostgreSQL release running on the server being
monitored.
"""
global latest_version_next_check
# If we don't know the latest version or it's past the recheck time, get the
# version from the PostgreSQL RSS feed. Only one thread needs to do this, so
# they all try to grab the lock, and then make sure nobody else beat them to it.
if (
latest_version is None
or latest_version_next_check is None
or latest_version_next_check < datetime.now()
Context.latest_version is None
or Context.latest_version_next_check is None
or Context.latest_version_next_check < datetime.now()
):
# Note: we get the cluster version here before grabbing the latest_version_lock
# lock so it's not held while trying to talk with the DB.
release = version_num_to_release(get_cluster_version())
with latest_version_lock:
with Context.latest_version_lock:
# Only check if nobody already got the version before us
if (
latest_version is None
or latest_version_next_check is None
or latest_version_next_check < datetime.now()
Context.latest_version is None
or Context.latest_version_next_check is None
or Context.latest_version_next_check < datetime.now()
):
log.info("Checking latest PostgreSQL version")
latest_version_next_check = datetime.now() + timedelta(
seconds=int(config["latest_version_check_period"])
Context.log.info("Checking latest PostgreSQL version")
Context.latest_version_next_check = datetime.now() + timedelta(
seconds=int(Context.config["latest_version_check_period"])
)
# Grab the RSS feed
raw_rss = requests.get("https://www.postgresql.org/versions.rss")
raw_rss = requests.get(
"https://www.postgresql.org/versions.rss", timeout=30
)
if raw_rss.status_code != 200:
raise LatestVersionCheckError("code={}".format(r.status_code))
raise LatestVersionCheckError("code={}".format(raw_rss.status_code))
# Parse the RSS body and set global variables
# Parse the RSS body and set Context variables
parse_version_rss(raw_rss.text, release)
return latest_version
return Context.latest_version
def sample_metric(dbname, metric_name, args, retry=True):
@@ -624,9 +730,9 @@ def sample_metric(dbname, metric_name, args, retry=True):
"""
# Get the metric definition
try:
metric = config["metrics"][metric_name]
except KeyError:
raise UnknownMetricError("Unknown metric: {}".format(metric_name))
metric = Context.config["metrics"][metric_name]
except KeyError as e:
raise UnknownMetricError("Unknown metric: {}".format(metric_name)) from e
# Get the connection pool for the database, or create one if it doesn't
# already exist.
@@ -641,8 +747,7 @@ def sample_metric(dbname, metric_name, args, retry=True):
# Execute the quert
if retry:
return run_query(pool, metric["type"], query, args)
else:
return run_query_no_retry(pool, metric["type"], query, args)
return run_query_no_retry(pool, metric["type"], query, args)
def test_queries():
@@ -650,12 +755,17 @@ def test_queries():
Run all of the metric queries against a database and check the results
"""
# We just use the default db for tests
dbname = config["dbname"]
dbname = Context.config["dbname"]
# Loop through all defined metrics.
for name, metric in config["metrics"].items():
for name, metric in Context.config["metrics"].items():
# If the metric has arguments to use while testing, grab those
args = metric.get("test_args", {})
print("Testing {} [{}]".format(name, ", ".join(["{}={}".format(key, value) for key, value in args.items()])))
print(
"Testing {} [{}]".format(
name,
", ".join(["{}={}".format(key, value) for key, value in args.items()]),
)
)
# When testing against a docker container, we may end up connecting
# before the service is truly up (it restarts during the initialization
# phase). To cope with this, we'll allow a few connection failures.
@@ -692,9 +802,8 @@ class SimpleHTTPRequestHandler(BaseHTTPRequestHandler):
"""
Override to suppress standard request logging
"""
pass
def do_GET(self):
def do_GET(self): # pylint: disable=invalid-name
"""
Handle a request. This is just a wrapper around the actual handler
code to keep things more readable.
@@ -702,7 +811,7 @@ class SimpleHTTPRequestHandler(BaseHTTPRequestHandler):
try:
self._handle_request()
except BrokenPipeError:
log.error("Client disconnected, exiting handler")
Context.log.error("Client disconnected, exiting handler")
def _handle_request(self):
"""
@@ -715,7 +824,6 @@ class SimpleHTTPRequestHandler(BaseHTTPRequestHandler):
if metric_name == "agent_version":
self._reply(200, VERSION)
return
elif metric_name == "latest_version_info":
try:
get_latest_version()
@@ -723,46 +831,40 @@ class SimpleHTTPRequestHandler(BaseHTTPRequestHandler):
200,
json.dumps(
{
"latest": latest_version,
"supported": 1 if release_supported else 0,
"latest": Context.latest_version,
"supported": 1 if Context.release_supported else 0,
}
),
)
except LatestVersionCheckError as e:
log.error("Failed to retrieve latest version information: {}".format(e))
Context.log.error(
"Failed to retrieve latest version information: %s", e
)
self._reply(503, "Failed to retrieve latest version info")
return
else:
# Note: parse_qs returns the values as a list. Since we always expect
# single values, just grab the first from each.
args = {key: values[0] for key, values in parsed_query.items()}
# Note: parse_qs returns the values as a list. Since we always expect
# single values, just grab the first from each.
args = {key: values[0] for key, values in parsed_query.items()}
# Get the dbname. If none was provided, use the default from the
# config.
dbname = args.get("dbname", Context.config["dbname"])
# Get the dbname. If none was provided, use the default from the
# config.
dbname = args.get("dbname", config["dbname"])
# Sample the metric
try:
self._reply(200, sample_metric(dbname, metric_name, args))
return
except UnknownMetricError as e:
log.error("Unknown metric: {}".format(metric_name))
self._reply(404, "Unknown metric")
return
except MetricVersionError as e:
log.error(
"Failed to find a version of {} for {}".format(metric_name, version)
)
self._reply(404, "Unsupported version")
return
except UnhappyDBError as e:
log.info("Database {} is unhappy, please be patient".format(dbname))
self._reply(503, "Database unavailable")
return
except Exception as e:
log.error("Error running query: {}".format(e))
self._reply(500, "Unexpected error: {}".format(e))
return
# Sample the metric
try:
self._reply(200, sample_metric(dbname, metric_name, args))
except UnknownMetricError:
Context.log.error("Unknown metric: %s", metric_name)
self._reply(404, "Unknown metric")
except MetricVersionError:
Context.log.error("Failed to find an query version for %s", metric_name)
self._reply(404, "Unsupported version")
except UnhappyDBError:
Context.log.info("Database %s is unhappy, please be patient", dbname)
self._reply(503, "Database unavailable")
except Exception as e: # pylint: disable=broad-exception-caught
Context.log.error("Error running query: %s", e)
self._reply(500, "Unexpected error: {}".format(e))
def _reply(self, code, content):
"""
@@ -775,7 +877,14 @@ class SimpleHTTPRequestHandler(BaseHTTPRequestHandler):
self.wfile.write(bytes(content, "utf-8"))
if __name__ == "__main__":
def main():
"""
Main application routine
"""
# Initialize the logging framework
Context.init_logging()
# Handle cli args
parser = argparse.ArgumentParser(
prog="pgmon", description="A PostgreSQL monitoring agent"
@@ -796,33 +905,35 @@ if __name__ == "__main__":
args = parser.parse_args()
# Set the config file path
config_file = args.config_file
Context.config_file = args.config_file
# Read the config file
read_config(config_file)
read_config(Context.config_file)
# Run query tests and exit if test mode is enabled
if args.test:
errors = test_queries()
if errors > 0:
if test_queries() > 0:
sys.exit(1)
else:
sys.exit(0)
sys.exit(0)
# Set up the http server to receive requests
server_address = (config["address"], config["port"])
httpd = ThreadingHTTPServer(server_address, SimpleHTTPRequestHandler)
server_address = (Context.config["address"], Context.config["port"])
Context.httpd = ThreadingHTTPServer(server_address, SimpleHTTPRequestHandler)
# Set up the signal handler
signal.signal(signal.SIGINT, signal_handler)
signal.signal(signal.SIGHUP, signal_handler)
# Handle requests.
log.info("Listening on port {}...".format(config["port"]))
while running:
httpd.handle_request()
Context.log.info("Listening on port %s...", Context.config["port"])
while Context.running:
Context.httpd.handle_request()
# Clean up PostgreSQL connections
# TODO: Improve this ... not sure it actually closes all the connections cleanly
for pool in connections.values():
for pool in Context.connections.values():
pool.close()
if __name__ == "__main__":
main()

File diff suppressed because it is too large Load Diff

129
zabbix_templates/coverage.py Executable file
View File

@@ -0,0 +1,129 @@
#!/usr/bin/env python3
# Compare the items defined in a Zabbix template with the metrics defined in a config file.
import sys
import yaml
# Special built in metrics
SPECIAL = {
"agent_version",
"latest_version_info"
}
class NonMetricItemError(Exception):
"""
A given item does not directly use a metric
"""
def read_metrics(file):
"""
Read the metrics from a config file and return the list of names
params:
file: the name of the file to read
returns:
list of metric named defined in the file
raises:
yaml.parser.ParserError: invalid yaml file
"""
names = set()
config = None
with open(file, "r", encoding="utf-8") as f:
config = yaml.safe_load(f)
try:
for m in config["metrics"].keys():
names.add(m)
except KeyError:
pass
return names
def extract_metric(item):
"""
Extract the metric from an item definition
params:
item: the item/discovery/prototype definition dict
returns:
the name of the metric used in the item
raises:
NonMetricItemError: the item does not directly use a metric
"""
try:
if item["type"] == "HTTP_AGENT":
url = item["url"]
if url.startswith("http://localhost:{$AGENT_PORT}"):
return url.split("/")[-1]
except KeyError:
raise NonMetricItemError()
raise NonMetricItemError()
def read_template(file):
"""
Read the items from a Zabbix template and return the list of metric names
params:
file: the name of the file to read
returns:
list of metric named used in the file
raises:
yaml.parser.ParserError: invalid yaml file
"""
names = set()
config = None
with open(file, "r", encoding="utf-8") as f:
config = yaml.safe_load(f)
try:
for template in config["zabbix_export"]["templates"]:
for item in template["items"]:
try:
names.add(extract_metric(item))
except NonMetricItemError:
pass
for rule in template["discovery_rules"]:
try:
names.add(extract_metric(rule))
except NonMetricItemError:
pass
for proto in rule["item_prototypes"]:
try:
names.add(extract_metric(proto))
except NonMetricItemError:
pass
except KeyError:
pass
return names
if __name__ == '__main__':
config_file = sys.argv[1]
config_metrics = read_metrics(config_file)
template_file = sys.argv[2]
template_metrics = read_template(template_file) - SPECIAL
config_only = config_metrics - template_metrics
template_only = template_metrics - config_metrics
print("Config only: {}".format(sorted(list(config_only))))
print("Template only: {}".format(sorted(list(template_only))))

View File

@@ -95,6 +95,20 @@ zabbix_export:
tags:
- tag: Application
value: PostgreSQL
- uuid: 06b1d082ed1e4796bc31cc25f7db6326
name: 'PostgreSQL Backend IO Info'
type: HTTP_AGENT
key: 'pgmon[io_per_backend]'
history: '0'
value_type: TEXT
trends: '0'
description: 'Aggregated statistics about I/O activity for different backend types'
url: 'http://localhost:{$AGENT_PORT}/io_per_backend'
tags:
- tag: Application
value: PostgreSQL
- tag: Type
value: Raw
- uuid: d890e395fbbc4f2bacbd50e7321fcb9f
name: 'PostgreSQL latest version info'
type: HTTP_AGENT
@@ -153,7 +167,8 @@ zabbix_export:
operator: NOT_MATCHES_REGEX
formulaid: A
lifetime: 30d
enabled_lifetime_type: DISABLE_NEVER
enabled_lifetime_type: DISABLE_AFTER
enabled_lifetime: 1d
item_prototypes:
- uuid: a30babe4a6f4440bba2a3ee46eff7ce2
name: 'Time spent executing statements on {#DBNAME}'
@@ -771,6 +786,77 @@ zabbix_export:
value: PostgreSQL
- tag: Database
value: '{#DBNAME}'
- uuid: 5960120dd01c4926b0fc1fbe9c011507
name: 'Database max sequence usage in {#DBNAME}'
type: HTTP_AGENT
key: 'pgmon_db_max_sequence[{#DBNAME}]'
delay: 5m
value_type: FLOAT
units: '%'
description: 'The percent of the currently configured value range for the most utilized sequence.'
url: 'http://localhost:{$AGENT_PORT}/sequence_usage'
query_fields:
- name: dbname
value: '{#DBNAME}'
tags:
- tag: Application
value: PostgreSQL
- tag: Database
value: '{#DBNAME}'
- uuid: 48b9cc80ac4d4aee9e9f3a5d6f7d4a95
name: 'Total number of sequences on {#DBNAME}'
type: DEPENDENT
key: 'pgmon_db_sequences[total,{#DBNAME}]'
delay: '0'
description: 'Total number of sequences in the database.'
preprocessing:
- type: JSONPATH
parameters:
- $.total_sequences
master_item:
key: 'pgmon_db_sequence_visibility[{#DBNAME}]'
tags:
- tag: Application
value: PostgreSQL
- tag: Database
value: '{#DBNAME}'
- uuid: 6521a9bab2ac47bf85429832d289bbac
name: 'Visible sequences on {#DBNAME}'
type: DEPENDENT
key: 'pgmon_db_sequences[visible,{#DBNAME}]'
delay: '0'
description: 'Number of sequences in the database for which Zabbix can see stats.'
preprocessing:
- type: JSONPATH
parameters:
- $.visible_sequences
master_item:
key: 'pgmon_db_sequence_visibility[{#DBNAME}]'
tags:
- tag: Application
value: PostgreSQL
- tag: Database
value: '{#DBNAME}'
- uuid: 00f2da3eb99940839410a6ecd5df153f
name: 'Database sequence visibility in {#DBNAME}'
type: HTTP_AGENT
key: 'pgmon_db_sequence_visibility[{#DBNAME}]'
delay: 30m
history: '0'
value_type: TEXT
trends: '0'
description: 'Statistics about the number of sequences that exist and the number Zabbix can actually see stats for.'
url: 'http://localhost:{$AGENT_PORT}/sequence_visibility'
query_fields:
- name: dbname
value: '{#DBNAME}'
tags:
- tag: Application
value: PostgreSQL
- tag: Database
value: '{#DBNAME}'
- tag: Type
value: Raw
- uuid: 492b3cac15f348c2b85f97b69c114d1b
name: 'Database Stats for {#DBNAME}'
type: HTTP_AGENT
@@ -789,6 +875,17 @@ zabbix_export:
value: '{#DBNAME}'
- tag: Type
value: Raw
trigger_prototypes:
- uuid: d29d0fd9d9d34b5ebd649592b0829ce5
expression: 'last(/PostgreSQL by pgmon/pgmon_db_sequences[total,{#DBNAME}]) <> last(/PostgreSQL by pgmon/pgmon_db_sequences[visible,{#DBNAME}])'
name: 'Sequences not visible to Zabbix on {#DBNAME}'
priority: WARNING
description: 'There are sequences for which Zabbix cannot see usage statistics'
tags:
- tag: Application
value: PostgreSQL
- tag: Component
value: Sequence
graph_prototypes:
- uuid: 1f7de43b77714f819e61c31273712b70
name: 'DML Totals for {#DBNAME}'
@@ -881,13 +978,613 @@ zabbix_export:
lld_macro_paths:
- lld_macro: '{#DBNAME}'
path: $.dbname
- uuid: 6afbe12fb9f54fa2ad29d647429eb16e
name: 'Discover I/O Backend Types'
type: DEPENDENT
key: pgmon_discover_io_backend_types
delay: '0'
lifetime: 30d
enabled_lifetime_type: DISABLE_AFTER
enabled_lifetime: 1h
item_prototypes:
- uuid: b1ac2e56b30f4812bf33ce973ef16b10
name: 'I/O Evictions by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend[evictions,{#BACKEND_TYPE}]'
delay: '0'
preprocessing:
- type: JSONPATH
parameters:
- '$[?(@.backend_type == "{#BACKEND_TYPE}")].evictions.first()'
master_item:
key: 'pgmon[io_per_backend]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: bde8e6c28e714be58fa963a395e14619
name: 'I/O Extends by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend[extends,{#BACKEND_TYPE}]'
delay: '0'
units: B
preprocessing:
- type: JSONPATH
parameters:
- '$[?(@.backend_type == "{#BACKEND_TYPE}")].extends.first()'
master_item:
key: 'pgmon[io_per_backend]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 5ce7a815c19241a6a398e17539cca1d4
name: 'I/O fsyncs by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend[fsyncs,{#BACKEND_TYPE}]'
delay: '0'
preprocessing:
- type: JSONPATH
parameters:
- '$[?(@.backend_type == "{#BACKEND_TYPE}")].fsyncs.first()'
master_item:
key: 'pgmon[io_per_backend]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 3be8cb08b6bc42f7888d8d7877ab759f
name: 'I/O Hits by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend[hits,{#BACKEND_TYPE}]'
delay: '0'
preprocessing:
- type: JSONPATH
parameters:
- '$[?(@.backend_type == "{#BACKEND_TYPE}")].hits.first()'
master_item:
key: 'pgmon[io_per_backend]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: e28566682196415fb3bdad9d9a294ae9
name: 'I/O Reads by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend[reads,{#BACKEND_TYPE}]'
delay: '0'
units: B
preprocessing:
- type: JSONPATH
parameters:
- '$[?(@.backend_type == "{#BACKEND_TYPE}")].reads.first()'
master_item:
key: 'pgmon[io_per_backend]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: e5e9b21c7f7649b284a8ca5b4502cb96
name: 'I/O Reuses by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend[reuses,{#BACKEND_TYPE}]'
delay: '0'
preprocessing:
- type: JSONPATH
parameters:
- '$[?(@.backend_type == "{#BACKEND_TYPE}")].reuses.first()'
master_item:
key: 'pgmon[io_per_backend]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 6de5bc1a98dc4d8da23ac0a90983c8aa
name: 'I/O Writebacks by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend[writebacks,{#BACKEND_TYPE}]'
delay: '0'
units: B
preprocessing:
- type: JSONPATH
parameters:
- '$[?(@.backend_type == "{#BACKEND_TYPE}")].writebacks.first()'
master_item:
key: 'pgmon[io_per_backend]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 87d09d3891c145a898050d390ea7f16e
name: 'I/O Writes by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend[writes,{#BACKEND_TYPE}]'
delay: '0'
units: B
preprocessing:
- type: JSONPATH
parameters:
- '$[?(@.backend_type == "{#BACKEND_TYPE}")].writes.first()'
master_item:
key: 'pgmon[io_per_backend]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 1587592caf694d91ba9457cafaca7761
name: 'I/O Evictions by {#BACKEND_TYPE} - 1h delta'
type: CALCULATED
key: 'pgmon_io_backend_delta[evictions,1h,{#BACKEND_TYPE}]'
delay: 10m
params: 'last(//pgmon_io_backend[evictions,{#BACKEND_TYPE}]) - last(//pgmon_io_backend[evictions,{#BACKEND_TYPE}], #1:now-1h)'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 637b0601458147b5b72bc40ab22b2117
name: 'I/O Evictions Rate by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend_delta[evictions,{#BACKEND_TYPE}]'
delay: '0'
value_type: FLOAT
preprocessing:
- type: CHANGE_PER_SECOND
parameters:
- ''
master_item:
key: 'pgmon_io_backend[evictions,{#BACKEND_TYPE}]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: c7ed8e1bcb0945ffa2a36b06ea5fe0c5
name: 'I/O Extends by {#BACKEND_TYPE} - 1h delta'
type: CALCULATED
key: 'pgmon_io_backend_delta[extends,1h,{#BACKEND_TYPE}]'
delay: 10m
units: B/h
params: 'last(//pgmon_io_backend[extends,{#BACKEND_TYPE}]) - last(//pgmon_io_backend[extends,{#BACKEND_TYPE}], #1:now-1h)'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 855cbbc83d884b00a6fea457f8e09806
name: 'I/O Extends Rate by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend_delta[extends,{#BACKEND_TYPE}]'
delay: '0'
value_type: FLOAT
units: Bps
preprocessing:
- type: CHANGE_PER_SECOND
parameters:
- ''
master_item:
key: 'pgmon_io_backend[extends,{#BACKEND_TYPE}]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 20c6135bb6c8491b9ca68fa506232b04
name: 'I/O fsyncs by {#BACKEND_TYPE} - 1h delta'
type: CALCULATED
key: 'pgmon_io_backend_delta[fsyncs,1h,{#BACKEND_TYPE}]'
delay: 10m
params: 'last(//pgmon_io_backend[fsyncs,{#BACKEND_TYPE}]) - last(//pgmon_io_backend[fsyncs,{#BACKEND_TYPE}], #1:now-1h)'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 7391c372d1704f21b314379c2fe5bb64
name: 'I/O fsyncs Rate by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend_delta[fsyncs,{#BACKEND_TYPE}]'
delay: '0'
value_type: FLOAT
preprocessing:
- type: CHANGE_PER_SECOND
parameters:
- ''
master_item:
key: 'pgmon_io_backend[fsyncs,{#BACKEND_TYPE}]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: ab8d0766bbf64db3bb09f9442d268a83
name: 'I/O Hits by {#BACKEND_TYPE} - 1h delta'
type: CALCULATED
key: 'pgmon_io_backend_delta[hits,1h,{#BACKEND_TYPE}]'
delay: 10m
params: 'last(//pgmon_io_backend[hits,{#BACKEND_TYPE}]) - last(//pgmon_io_backend[hits,{#BACKEND_TYPE}], #1:now-1h)'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: be9aa056b9364c12a8809d5dd3af8387
name: 'I/O Hits Rate by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend_delta[hits,{#BACKEND_TYPE}]'
delay: '0'
value_type: FLOAT
preprocessing:
- type: CHANGE_PER_SECOND
parameters:
- ''
master_item:
key: 'pgmon_io_backend[hits,{#BACKEND_TYPE}]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 0c42a40a263948dda4464856747ac6ef
name: 'I/O Reads by {#BACKEND_TYPE} - 1h delta'
type: CALCULATED
key: 'pgmon_io_backend_delta[reads,1h,{#BACKEND_TYPE}]'
delay: 10m
units: B/h
params: 'last(//pgmon_io_backend[reads,{#BACKEND_TYPE}]) - last(//pgmon_io_backend[reads,{#BACKEND_TYPE}], #1:now-1h)'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 01bc131aaec448e6aafb7ef13d1ea5ed
name: 'I/O Reads Rate by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend_delta[reads,{#BACKEND_TYPE}]'
delay: '0'
value_type: FLOAT
units: Bps
preprocessing:
- type: CHANGE_PER_SECOND
parameters:
- ''
master_item:
key: 'pgmon_io_backend[reads,{#BACKEND_TYPE}]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 740d5c84be6e4b99b4ca3513b6c6c5ea
name: 'I/O Reuses by {#BACKEND_TYPE} - 1h delta'
type: CALCULATED
key: 'pgmon_io_backend_delta[reuses,1h,{#BACKEND_TYPE}]'
delay: 10m
params: 'last(//pgmon_io_backend[reuses,{#BACKEND_TYPE}]) - last(//pgmon_io_backend[reuses,{#BACKEND_TYPE}], #1:now-1h)'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 8cbc16fc472a44d5945498285e8407be
name: 'I/O Reuses Rate by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend_delta[reuses,{#BACKEND_TYPE}]'
delay: '0'
value_type: FLOAT
preprocessing:
- type: CHANGE_PER_SECOND
parameters:
- ''
master_item:
key: 'pgmon_io_backend[reuses,{#BACKEND_TYPE}]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 2e796f4c75a94967a0b11a37a2d64726
name: 'I/O Writebacks by {#BACKEND_TYPE} - 1h delta'
type: CALCULATED
key: 'pgmon_io_backend_delta[writebacks,1h,{#BACKEND_TYPE}]'
delay: 10m
units: B/h
params: 'last(//pgmon_io_backend[writebacks,{#BACKEND_TYPE}]) - last(//pgmon_io_backend[writebacks,{#BACKEND_TYPE}], #1:now-1h)'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 52d2536db79047948893ccb764b2a220
name: 'I/O Writebacks Rate by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend_delta[writebacks,{#BACKEND_TYPE}]'
delay: '0'
value_type: FLOAT
units: Bps
preprocessing:
- type: CHANGE_PER_SECOND
parameters:
- ''
master_item:
key: 'pgmon_io_backend[writebacks,{#BACKEND_TYPE}]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 6a741d3b857b48eea0e1a4d1deea7183
name: 'I/O Writes by {#BACKEND_TYPE} - 1h delta'
type: CALCULATED
key: 'pgmon_io_backend_delta[writes,1h,{#BACKEND_TYPE}]'
delay: 10m
units: B/h
params: 'last(//pgmon_io_backend[writes,{#BACKEND_TYPE}]) - last(//pgmon_io_backend[writes,{#BACKEND_TYPE}], #1:now-1h)'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
- uuid: 6d10339b08b84a91a931a64050a471cc
name: 'I/O Writes Rate by {#BACKEND_TYPE}'
type: DEPENDENT
key: 'pgmon_io_backend_delta[writes,{#BACKEND_TYPE}]'
delay: '0'
value_type: FLOAT
units: Bps
preprocessing:
- type: CHANGE_PER_SECOND
parameters:
- ''
master_item:
key: 'pgmon_io_backend[writes,{#BACKEND_TYPE}]'
tags:
- tag: Application
value: PostgreSQL
- tag: 'Backend Type'
value: '{#BACKEND_TYPE}'
- tag: Component
value: IO
graph_prototypes:
- uuid: faf1e3f41dd84adba42c414f6efac152
name: 'Backend I/O for {#BACKEND_TYPE}'
graph_items:
- color: 199C0D
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend[evictions,{#BACKEND_TYPE}]'
- sortorder: '1'
color: F63100
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend[extends,{#BACKEND_TYPE}]'
- sortorder: '2'
color: 2774A4
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend[fsyncs,{#BACKEND_TYPE}]'
- sortorder: '3'
color: F7941D
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend[hits,{#BACKEND_TYPE}]'
- sortorder: '4'
color: FC6EA3
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend[reads,{#BACKEND_TYPE}]'
- sortorder: '5'
color: 6C59DC
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend[reuses,{#BACKEND_TYPE}]'
- sortorder: '6'
color: C7A72D
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend[writebacks,{#BACKEND_TYPE}]'
- sortorder: '7'
color: BA2A5D
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend[writes,{#BACKEND_TYPE}]'
- uuid: 24e2e3ec67cd43058c8b6a02d686fc6a
name: 'Backend I/O for {#BACKEND_TYPE} - 1h delta'
graph_items:
- color: 199C0D
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[evictions,1h,{#BACKEND_TYPE}]'
- sortorder: '1'
color: F63100
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[extends,1h,{#BACKEND_TYPE}]'
- sortorder: '2'
color: 2774A4
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[fsyncs,1h,{#BACKEND_TYPE}]'
- sortorder: '3'
color: F7941D
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[hits,1h,{#BACKEND_TYPE}]'
- sortorder: '4'
color: FC6EA3
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[reads,1h,{#BACKEND_TYPE}]'
- sortorder: '5'
color: 6C59DC
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[reuses,1h,{#BACKEND_TYPE}]'
- sortorder: '6'
color: C7A72D
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[writebacks,1h,{#BACKEND_TYPE}]'
- sortorder: '7'
color: BA2A5D
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[writes,1h,{#BACKEND_TYPE}]'
- uuid: 815f1a00c19d45a18b10846644710668
name: 'Backend I/O Rates for {#BACKEND_TYPE}'
graph_items:
- color: 199C0D
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[evictions,{#BACKEND_TYPE}]'
- sortorder: '1'
color: F63100
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[extends,{#BACKEND_TYPE}]'
- sortorder: '2'
color: 2774A4
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[fsyncs,{#BACKEND_TYPE}]'
- sortorder: '3'
color: F7941D
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[hits,{#BACKEND_TYPE}]'
- sortorder: '4'
color: FC6EA3
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[reads,{#BACKEND_TYPE}]'
- sortorder: '5'
color: 6C59DC
yaxisside: RIGHT
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[reuses,{#BACKEND_TYPE}]'
- sortorder: '6'
color: C7A72D
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[writebacks,{#BACKEND_TYPE}]'
- sortorder: '7'
color: BA2A5D
calc_fnc: ALL
item:
host: 'PostgreSQL by pgmon'
key: 'pgmon_io_backend_delta[writes,{#BACKEND_TYPE}]'
master_item:
key: 'pgmon[io_per_backend]'
lld_macro_paths:
- lld_macro: '{#BACKEND_TYPE}'
path: $.backend_type
preprocessing:
- type: JSONPATH
parameters:
- '$[:]'
- uuid: 8ec029d577ae4872858e2e5cfd1cc40e
name: 'Discover Replication'
type: HTTP_AGENT
key: pgmon_discover_rep
delay: 10m
filter:
conditions:
- macro: '{#APPLICATION_NAME}'
value: '^pg_[0-9]+_sync_[0-9]+_[0-9]+$'
operator: NOT_MATCHES_REGEX
formulaid: A
lifetime: 30d
enabled_lifetime_type: DISABLE_NEVER
enabled_lifetime_type: DISABLE_AFTER
enabled_lifetime: 7d
item_prototypes:
- uuid: 3a5a60620e6a4db694e47251148d82f5
name: 'Flush lag for {#REPID}'
@@ -896,6 +1593,7 @@ zabbix_export:
delay: '0'
history: 90d
value_type: FLOAT
units: s
description: 'Time elapsed between flushing recent WAL locally and receiving notification that this standby server has written and flushed it (but not yet applied it). This can be used to gauge the delay that synchronous_commit level on incurred while committing if this server was configured as a synchronous standby.'
preprocessing:
- type: JSONPATH
@@ -911,13 +1609,12 @@ zabbix_export:
- tag: Database
value: '{#DBNAME}'
- uuid: 624f8f085a3642c9a10a03361c17763d
name: 'Last flush LSN for {#REPID}'
name: 'Last flush LSN lag for {#REPID}'
type: DEPENDENT
key: 'pgmon_rep[flush_lsn,repid={#REPID}]'
delay: '0'
history: 90d
value_type: TEXT
trends: '0'
units: B
description: 'Last write-ahead log location flushed to disk by this standby server'
preprocessing:
- type: JSONPATH
@@ -939,6 +1636,7 @@ zabbix_export:
delay: '0'
history: 90d
value_type: FLOAT
units: s
description: 'Time elapsed between flushing recent WAL locally and receiving notification that this standby server has written, flushed and applied it. This can be used to gauge the delay that synchronous_commit level remote_apply incurred while committing if this server was configured as a synchronous standby.'
preprocessing:
- type: JSONPATH
@@ -954,13 +1652,12 @@ zabbix_export:
- tag: Database
value: '{#DBNAME}'
- uuid: fe1bed51845d4694bae8f53deed4846d
name: 'Last replay LSN for {#REPID}'
name: 'Last replay LSN lag for {#REPID}'
type: DEPENDENT
key: 'pgmon_rep[replay_lsn,repid={#REPID}]'
delay: '0'
history: 90d
value_type: TEXT
trends: '0'
units: B
description: 'Last write-ahead log location replayed into the database on this standby server'
preprocessing:
- type: JSONPATH
@@ -976,13 +1673,12 @@ zabbix_export:
- tag: Database
value: '{#DBNAME}'
- uuid: 68c179d0e33f45f9bf82d2d4125763f0
name: 'Last sent LSN for {#REPID}'
name: 'Last sent LSN lag for {#REPID}'
type: DEPENDENT
key: 'pgmon_rep[sent_lsn,repid={#REPID}]'
delay: '0'
history: 90d
value_type: TEXT
trends: '0'
units: B
description: 'Last write-ahead log location sent on this connection'
preprocessing:
- type: JSONPATH
@@ -1032,6 +1728,7 @@ zabbix_export:
delay: '0'
history: 90d
value_type: FLOAT
units: s
description: 'Time elapsed between flushing recent WAL locally and receiving notification that this standby server has written it (but not yet flushed it or applied it). This can be used to gauge the delay that synchronous_commit level remote_write incurred while committing if this server was configured as a synchronous standby.'
preprocessing:
- type: JSONPATH
@@ -1047,13 +1744,12 @@ zabbix_export:
- tag: Database
value: '{#DBNAME}'
- uuid: 57fb03cf63af4b0a91d8e36d6ff64d30
name: 'Last write LSN for {#REPID}'
name: 'Last write LSN lag for {#REPID}'
type: DEPENDENT
key: 'pgmon_rep[write_lsn,repid={#REPID}]'
delay: '0'
history: 90d
value_type: TEXT
trends: '0'
units: B
description: 'Time elapsed between flushing recent WAL locally and receiving notification that this standby server has written it (but not yet flushed it or applied it). This can be used to gauge the delay that synchronous_commit level remote_write incurred while committing if this server was configured as a synchronous standby.'
preprocessing:
- type: JSONPATH
@@ -1090,6 +1786,8 @@ zabbix_export:
value: Raw
url: 'http://localhost:{$AGENT_PORT}/discover_rep'
lld_macro_paths:
- lld_macro: '{#APPLICATION_NAME}'
path: $.application_name
- lld_macro: '{#CLIENT_ADDR}'
path: $.client_addr
- lld_macro: '{#REPID}'
@@ -1101,6 +1799,15 @@ zabbix_export:
type: HTTP_AGENT
key: pgmon_discover_slots
delay: 10m
filter:
conditions:
- macro: '{#SLOT_NAME}'
value: '^pg_[0-9]+_sync_[0-9]+_[0-9]+$'
operator: NOT_MATCHES_REGEX
formulaid: A
lifetime: 30d
enabled_lifetime_type: DISABLE_AFTER
enabled_lifetime: 7d
item_prototypes:
- uuid: 536c5f82e3074ddfbfd842b3a2e8d46c
name: 'Slot {#SLOT_NAME} - Confirmed Flushed Bytes Lag'