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.

.gitignore

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

DEBIAN/control

@@ -3,7 +3,7 @@ Version: 1.0
 Section: utils
 Priority: optional
 Architecture: all
-Depends: logrotate, python3 (>= 3.6), python3-psycopg2, python3-yaml, systemd
+Depends: logrotate, python3 (>= 3.6), python3-psycopg2, python3-requests, python3-yaml, systemd
 Maintainer: James Campbell <james@commandprompt.com>
 Homepage: https://www.commandprompt.com
 Description: A bridge to sit between monitoring tools and PostgreSQL

GENTOO/pgmon-1.0.3.ebuild

@@ -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
+}

GENTOO/pgmon-1.0.4.ebuild

@@ -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
+}

Makefile

@@ -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
+.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 $(BUILD_DIR) .
+	tar --transform "s,^,$(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
@@ -54,33 +79,53 @@ package-%:
 tgz:
 	rm -rf $(BUILD_DIR)/tgz/root
 	mkdir -p $(BUILD_DIR)/tgz/root
-	$(MAKE) install DESTDIR=$(BUILD_DIR)/tgz/root
-	tar -cz -f $(BUILD_DIR)/tgz/$(PACKAGE_NAME)-$(VERSION).tgz -C $(BUILD_DIR)/tgz/root .
+	$(MAKE) install-openrc DESTDIR=$(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
 
-# Install the script at the specified base directory
-install:
+# 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
 	mkdir -p $(DESTDIR)/etc/$(PACKAGE_NAME)
 	mkdir -p ${DESTDIR}/etc/logrotate.d
-	mkdir -p $(DESTDIR)/lib/systemd/system
-	mkdir -p $(DESTDIR)/usr/local/bin
+	mkdir -p $(DESTDIR)/usr/bin
 	mkdir -p $(DESTDIR)/usr/share/man/man1
 	
 	# Install script
-	cp $(SCRIPT) $(DESTDIR)/usr/local/bin/$(PACKAGE_NAME)
-	chmod 755 $(DESTDIR)/usr/local/bin/$(PACKAGE_NAME)
+	cp $(SCRIPT) $(DESTDIR)/usr/bin/$(PACKAGE_NAME)
+	chmod 755 $(DESTDIR)/usr/bin/$(PACKAGE_NAME)
 	
 	# Install manpage
 	cp manpages/* $(DESTDIR)/usr/share/man/man1/
@@ -89,12 +134,36 @@ install:
 	# Install sample config
 	cp sample-config/* $(DESTDIR)/etc/$(PACKAGE_NAME)/
 	
-	# Install systemd unit files
-	cp systemd/* $(DESTDIR)/lib/systemd/system/
-	
 	# Install logrotate config
 	cp logrotate/${PACKAGE_NAME}.logrotate ${DESTDIR}/etc/logrotate.d/${PACKAGE_NAME}
 
+# Install for systemd
+install-systemd:
+	# Install the common stuff
+	$(MAKE) install-common
+	
+	# Set up directories
+	mkdir -p $(DESTDIR)/lib/systemd/system
+	
+	# Install systemd unit files
+	cp systemd/* $(DESTDIR)/lib/systemd/system/
+
+# Install for open-rc
+install-openrc:
+	# Install the common stuff
+	$(MAKE) install-common
+	
+	# Set up directories
+	mkdir -p $(DESTDIR)/etc/init.d
+	mkdir -p $(DESTDIR)/etc/conf.d
+	
+	# Install init script
+	cp openrc/pgmon.initd $(DESTDIR)/etc/init.d/pgmon
+	chmod 755 $(DESTDIR)/etc/init.d/pgmon
+	
+	# Install init script config file
+	cp openrc/pgmon.confd $(DESTDIR)/etc/conf.d/pgmon
+
 
 # Run all of the install tests
 .PHONY: install-tests debian-%-install-test rockylinux-%-install-test ubuntu-%-install-test gentoo-install-test
@@ -106,28 +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:
@@ -167,30 +236,30 @@ package-image-%:
 
 # Debian package creation
 actually-package-debian-%:
-	$(MAKE) install DESTDIR=/output/debian-$*
+	$(MAKE) install-systemd DESTDIR=/output/debian-$*
 	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-%:
-	$(MAKE) install DESTDIR=/output/ubuntu-$*
+	$(MAKE) install-systemd DESTDIR=/output/ubuntu-$*
 	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/

RPM/pgmon-el7.spec

@@ -1,13 +1,13 @@
 Name: pgmon
 Version: @@VERSION@@
-Release: 1%{?dist}
+Release: @@RELEASE@@%{?dist}
 Summary: A bridge to sit between monitoring tools and PostgreSQL
 
 License: MIT
 URL: https://www.commandprompt.com
 
 BuildArch: noarch
-Requires: logrotate, python, python-psycopg2, PyYAML, systemd
+Requires: logrotate, python, python-psycopg2, PyYAML, python-requests, systemd
 
 %description
 A bridge to sit between monitoring tools and PostgreSQL
@@ -19,7 +19,7 @@ A bridge to sit between monitoring tools and PostgreSQL
 # Do nothing since we have nothing to build
 
 %install
-make -C /src install DESTDIR=%{buildroot}
+make -C /src install-systemd DESTDIR=%{buildroot}
 
 %files
 /etc/logrotate.d/pgmon
@@ -28,7 +28,7 @@ make -C /src install DESTDIR=%{buildroot}
 /etc/pgmon/pgmon-service.conf
 /lib/systemd/system/pgmon.service
 /lib/systemd/system/pgmon@.service
-/usr/local/bin/pgmon
+/usr/bin/pgmon
 /usr/share/man/man1/pgmon.1.gz
 
 %post

RPM/pgmon.spec

@@ -1,13 +1,13 @@
 Name: pgmon
 Version: @@VERSION@@
-Release: 1%{?dist}
+Release: @@RELEASE@@%{?dist}
 Summary: A bridge to sit between monitoring tools and PostgreSQL
 
 License: MIT
 URL: https://www.commandprompt.com
 
 BuildArch: noarch
-Requires: logrotate, python3, python3-psycopg2, python3-pyyaml, systemd
+Requires: logrotate, python3, python3-psycopg2, python3-pyyaml, python3-requests, systemd
 
 %description
 A bridge to sit between monitoring tools and PostgreSQL
@@ -19,7 +19,7 @@ A bridge to sit between monitoring tools and PostgreSQL
 # Do nothing since we have nothing to build
 
 %install
-make -C /src install DESTDIR=%{buildroot}
+make -C /src install-systemd DESTDIR=%{buildroot}
 
 %files
 /etc/logrotate.d/pgmon
@@ -28,7 +28,7 @@ make -C /src install DESTDIR=%{buildroot}
 /etc/pgmon/pgmon-service.conf
 /lib/systemd/system/pgmon.service
 /lib/systemd/system/pgmon@.service
-/usr/local/bin/pgmon
+/usr/bin/pgmon
 /usr/share/man/man1/pgmon.1.gz
 
 %post

pylintrc

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

sample-config/pgmon-metrics.yml

@@ -1,63 +1,320 @@
 metrics:
+  ##
   # Discovery metrics
+  ##
   discover_dbs:
     type: set
     query:
-      0: SELECT datname AS dbname FROM pg_database
+      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, client_addr, state FROM pg_stat_replication
+      0: >
+        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:
-      90400: SELECT slot_name, plugin, slot_type, database, false as temporary, active FROM pg_replication_slots
-      100000: SELECT slot_name, plugin, slot_type, database, temporary, active FROM pg_replication_slots
+      90400: >
+        SELECT slot_name,
+               plugin,
+               slot_type,
+               database,
+               false as temporary,
+               active
+        FROM pg_replication_slots
+      100000: >
+        SELECT slot_name,
+               plugin,
+               slot_type,
+               database,
+               temporary,
+               active
+        FROM pg_replication_slots
 
+
+  ##
   # cluster-wide metrics
+  ##
   version:
     type: value
     query:
       0: SHOW server_version_num
+
   max_frozen_age:
     type: row
     query:
-      0: SELECT max(age(datfrozenxid)), max(mxid_age(datminmxid)) FROM pg_database
+      0: >
+        SELECT max(age(datfrozenxid)) AS xid_age,
+               NULL AS mxid_age
+        FROM pg_database
+      90600: >
+        SELECT max(age(datfrozenxid)) AS xid_age,
+               max(mxid_age(datminmxid)) AS mxid_age
+        FROM pg_database
 
+  bgwriter:
+    type: row
+    query:
+      0: >
+        SELECT checkpoints_timed,
+               checkpoints_req,
+               checkpoint_write_time,
+               checkpoint_sync_time,
+               buffers_checkpoint,
+               buffers_clean,
+               maxwritten_clean,
+               buffers_backend,
+               buffers_backend_fsync,
+               buffers_alloc
+        FROM pg_stat_bgwriter
+      170000: >
+        SELECT cp.num_timed AS checkpoints_timed,
+               cp.num_requested AS checkpoints_req,
+               cp.write_time AS checkpoint_write_time,
+               cp.sync_time AS checkpoint_sync_time,
+               cp.buffers_written AS buffers_checkpoint,
+               bg.buffers_clean AS buffers_clean,
+               bg.maxwritten_clean AS maxwritten_clean,
+               NULL AS buffers_backend,
+               NULL AS buffers_backend_fsync,
+               bg.buffers_alloc AS buffers_alloc
+        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
+  ##
   db_stats:
     type: row
     query:
-      0: SELECT numbackends, xact_commit, xact_rollback, blks_read, blks_hit, tup_returned, tup_fetched, tup_inserted, tup_updated, tup_deleted, conflicts, temp_files, temp_bytes, deadlocks, blk_read_time, blk_write_time, extract('epoch' from stats_reset)::float FROM pg_stat_database WHERE datname = %(dbname)s
-      140000: SELECT numbackends, xact_commit, xact_rollback, blks_read, blks_hit, tup_returned, tup_fetched, tup_inserted, tup_updated, tup_deleted, conflicts, temp_files, temp_bytes, deadlocks, COALESCE(checksum_failures, 0) AS checksum_failures, blk_read_time, blk_write_time, session_time, active_time, idle_in_transaction_time, sessions, sessions_abandoned, sessions_fatal, sessions_killed, extract('epoch' from stats_reset)::float FROM pg_stat_database WHERE datname = %(dbname)s
+      0: >
+        SELECT numbackends,
+               xact_commit,
+               xact_rollback,
+               blks_read,
+               blks_hit,
+               tup_returned,
+               tup_fetched,
+               tup_inserted,
+               tup_updated,
+               tup_deleted,
+               conflicts,
+               temp_files,
+               temp_bytes,
+               deadlocks,
+               NULL AS checksum_failures,
+               blk_read_time,
+               blk_write_time,
+               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,
+               xact_commit,
+               xact_rollback,
+               blks_read,
+               blks_hit,
+               tup_returned,
+               tup_fetched,
+               tup_inserted,
+               tup_updated,
+               tup_deleted,
+               conflicts,
+               temp_files,
+               temp_bytes,
+               deadlocks,
+               COALESCE(checksum_failures, 0) AS checksum_failures,
+               blk_read_time,
+               blk_write_time,
+               session_time,
+               active_time,
+               idle_in_transaction_time,
+               sessions,
+               sessions_abandoned,
+               sessions_fatal,
+               sessions_killed,
+               extract('epoch' from stats_reset) AS stats_reset
+        FROM pg_stat_database WHERE datname = %(dbname)s
     test_args:
       dbname: postgres
 
+  hit_ratios:
+    type: row
+    query:
+      0: >
+        SELECT sum(heap_blks_read)::float / NULLIF(sum(heap_blks_read + heap_blks_hit), 0) AS avg_heap_hit_ratio,
+               sum(idx_blks_hit)::float / NULLIF(sum(idx_blks_read + idx_blks_hit), 0) AS avg_idx_hit_ratio,
+               sum(toast_blks_hit)::float / NULLIF(sum(toast_blks_read + toast_blks_hit), 0) AS avg_toast_hit_ratio,
+               sum(tidx_blks_hit)::float / NULLIF(sum(tidx_blks_read + tidx_blks_hit), 0) AS avg_tidx_hit_ratio
+        FROM pg_statio_all_tables
+    test_args:
+      dbname: postgres
+
+  activity:
+    type: set
+    query:
+      0: >
+        SELECT state,
+               count(*) AS backend_count,
+               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
+  ##
+  rep_stats:
+    type: row
+    query:
+      90400: >
+        SELECT pid, usename,
+               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,
+               pg_xlog_location_diff(pg_current_xlog_location(), flush_location) AS flush_lsn,
+               pg_xlog_location_diff(pg_current_xlog_location(), replay_location) AS replay_lsn,
+               NULL AS write_lag,
+               NULL AS flush_lag,
+               NULL AS replay_lag,
+               sync_state
+        FROM pg_stat_replication
+        WHERE host(client_addr) || '_' || regexp_replace(application_name, '[ ,]', '_', 'g') = %(repid)s
+      100000: >
+        SELECT pid, usename,
+               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) 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 host(client_addr) || '_' || regexp_replace(application_name, '[ ,]', '_', 'g') = %(repid)s
+    test_args:
+      repid: 127.0.0.1_test_rep
+
+
+  ##
+  # Per-slot metrics
+  ##
+  slot_stats:
+    type: row
+    query:
+      90400: >
+        SELECT NULL as active_pid,
+               xmin,
+               pg_xlog_location_diff(pg_current_xlog_location(), restart_lsn) AS restart_bytes,
+               NULL AS confirmed_flush_bytes
+        FROM pg_replication_slots WHERE slot_name = %(slot)s
+      90600: >
+        SELECT active_pid,
+               xmin,
+               pg_xlog_location_diff(pg_current_xlog_location(), restart_lsn) AS restart_bytes,
+               pg_xlog_location_diff(pg_current_xlog_location(), confirmed_flush_lsn) AS confirmed_flush_bytes
+        FROM pg_replication_slots WHERE slot_name = %(slot)s
+      100000: >
+        SELECT active_pid,
+               xmin,
+               pg_wal_lsn_diff(pg_current_wal_lsn(), restart_lsn) AS restart_bytes,
+               pg_wal_lsn_diff(pg_current_wal_lsn(), confirmed_flush_lsn) AS confirmed_flush_bytes
+        FROM pg_replication_slots WHERE slot_name = %(slot)s
+    test_args:
+      slot: test_slot
+
+
+  ##
   # Debugging
+  ##
   ntables:
     type: value
     query:
       0: SELECT count(*) AS ntables FROM pg_stat_user_tables
-
-  # Per-replication metrics
-  rep_stats:
-    type: row
-    query:
-      90400: SELECT * FROM pg_stat_replication WHERE client_addr || '_' || regexp_replace(application_name, '[ ,]', '_', 'g') = '{repid}'
-    test_args:
-      repid: 127.0.0.1_test_rep
-
-  # Debugging
-  sleep:
-    type: value
-    query:
-      0: SELECT now(), pg_sleep(5);
-
-  # Per-slot metrics
-  slot_stats:
-    type: row
-    query:
-      90400: SELECT active_pid, xmin, pg_xlog_location_diff(pg_current_xlog_location(), restart_lsn) AS restart_bytes, pg_xlog_location_diff(pg_current_xlog_location(), confirmed_flush_lsn) AS confirmed_flush_bytes FROM pg_replication_slots WHERE slot_name = '{slot}'
-      100000: SELECT active_pid, xmin, pg_wal_lsn_diff(pg_current_wal_lsn(), restart_lsn) AS restart_bytes, pg_wal_lsn_diff(pg_current_wal_lsn(), confirmed_flush_lsn) AS confirmed_flush_bytes FROM pg_replication_slots WHERE slot_name = '{slot}'
-    test_args:
-      slot: test_slot

src/pgmon.py

@@ -1,90 +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 requests
 
-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
+VERSION = "1.0.4"
 
-VERSION = "1.0.2"
 
-# Configuration
-config = {}
+class Context:
+    """
+    The global context for connections, config, version, nad IPC
+    """
 
-# Dictionary of current PostgreSQL connection pools
-connections_lock = Lock()
-connections = {}
+    # Configuration
+    config = {}
 
-# 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 = {}
+    # Dictionary of current PostgreSQL connection pools
+    connections_lock = Lock()
+    connections = {}
 
-# Version information
-cluster_version = None
-cluster_version_next_check = None
-cluster_version_lock = Lock()
+    # 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 = {}
 
-# Running state (used to gracefully shut down)
-running = True
+    # Version information
+    cluster_version = None
+    cluster_version_next_check = None
+    cluster_version_lock = Lock()
 
-# The http server object
-httpd = None
+    # PostgreSQL latest version information
+    latest_version = None
+    latest_version_next_check = None
+    latest_version_lock = Lock()
+    release_supported = None
 
-# Where the config file lives
-config_file = None
+    # Running state (used to gracefully shut down)
+    running = True
 
-# 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)
+    # 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):
+    """
+    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
@@ -116,6 +173,8 @@ default_config = {
     "reconnect_cooldown": 30,
     # How often to check the version of PostgreSQL (seconds)
     "version_check_period": 300,
+    # How often to check the latest supported version of PostgreSQL (seconds)
+    "latest_version_check_period": 86400,
     # Metrics
     "metrics": {},
 }
@@ -160,6 +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.
@@ -168,61 +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("/"):
@@ -233,34 +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
 
@@ -271,19 +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
@@ -292,7 +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
@@ -316,34 +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):
@@ -351,8 +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"])
     )
 
 
@@ -376,38 +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):
@@ -428,7 +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
@@ -436,9 +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():
@@ -446,40 +571,157 @@ 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):
+    """
+    Extract the revease from a version_num.
+
+    In other words, this converts things like:
+        90603  => 9.6
+        130010 => 13
+    """
+    if version_num // 10000 < 10:
+        return version_num // 10000 + (version_num % 10000 // 100 / 10)
+    return version_num // 10000
+
+
+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 Context variables:
+        latest_version
+        release_supported
+
+    It is expected that the caller already holds the latest_version_lock lock.
+
+    params:
+        raw_rss: The raw rss text from versions.rss
+        release: The PostgreSQL release we care about (ex: 9.2, 14)
+    """
+
+    # Regular expressions for parsing the RSS document
+    version_line = re.compile(
+        r".*?([0-9][0-9.]+) is the latest release in the {} series.*".format(release)
+    )
+    unsupported_line = re.compile(r"^This version is unsupported")
+
+    # Loop through the RSS until we find the current release
+    release_found = False
+    for line in raw_rss.splitlines():
+        m = version_line.match(line)
+        if m:
+            # Note that we found the version we were looking for
+            release_found = True
+
+            # Convert the version to version_num format
+            version = m.group(1)
+            parts = list(map(int, version.split(".")))
+            if parts[0] < 10:
+                Context.latest_version = int(
+                    "{}{:02}{:02}".format(parts[0], parts[1], parts[2])
+                )
+            else:
+                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):
+                Context.release_supported = False
+            else:
+                Context.release_supported = True
+            break
+
+    # Make sure we actually found it
+    if not release_found:
+        raise LatestVersionCheckError("Current release ({}) not found".format(release))
+
+    Context.log.info(
+        "Got latest PostgreSQL version: %s supported=%s",
+        Context.latest_version,
+        Context.release_supported,
+    )
+    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.
+    """
+
+    # 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 (
+        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 Context.latest_version_lock:
+            # Only check if nobody already got the version before us
+            if (
+                Context.latest_version is None
+                or Context.latest_version_next_check is None
+                or Context.latest_version_next_check < datetime.now()
+            ):
+                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", timeout=30
+                )
+                if raw_rss.status_code != 200:
+                    raise LatestVersionCheckError("code={}".format(raw_rss.status_code))
+
+                # Parse the RSS body and set Context variables
+                parse_version_rss(raw_rss.text, release)
+
+    return Context.latest_version
 
 
 def sample_metric(dbname, metric_name, args, retry=True):
@@ -488,9 +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.
@@ -505,8 +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():
@@ -514,13 +755,35 @@ 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", {})
-        # Run the query without the ability to retry.
-        res = sample_metric(dbname, name, args, retry=False)
+        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.
+        tries = 5
+        while True:
+            # Run the query without the ability to retry
+            try:
+                res = sample_metric(dbname, name, args, retry=False)
+                break
+            except MetricVersionError:
+                res = "Unsupported for this version"
+                break
+            except psycopg2.OperationalError as e:
+                print("Error encountered, {} tries left: {}".format(tries, e))
+                if tries <= 0:
+                    raise
+                time.sleep(1)
+            tries -= 1
         # Compare the result to the provided sample results
         # TODO
         print("{} -> {}".format(name, res))
@@ -539,9 +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.
@@ -549,7 +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):
         """
@@ -562,38 +824,47 @@ class SimpleHTTPRequestHandler(BaseHTTPRequestHandler):
 
         if metric_name == "agent_version":
             self._reply(200, VERSION)
-            return
+        elif metric_name == "latest_version_info":
+            try:
+                get_latest_version()
+                self._reply(
+                    200,
+                    json.dumps(
+                        {
+                            "latest": Context.latest_version,
+                            "supported": 1 if Context.release_supported else 0,
+                        }
+                    ),
+                )
+            except LatestVersionCheckError as e:
+                Context.log.error(
+                    "Failed to retrieve latest version information: %s", e
+                )
+                self._reply(503, "Failed to retrieve latest version info")
+        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):
         """
@@ -606,7 +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"
@@ -627,33 +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()

systemd/pgmon@.service

@@ -7,7 +7,7 @@ After=network.target
 [Service]
 EnvironmentFile=/etc/pgmon/%i-service.conf
 User=${SERVICE_USER:-postgres}
-ExecStart=/usr/local/bin/pgmon -c /etc/pgmon/%i.yml
+ExecStart=/usr/bin/pgmon -c /etc/pgmon/%i.yml
 ExecReload=kill -HUP $MAINPID
 Restart=on-failure
 Type=exec

tests/Dockerfile

@@ -2,6 +2,7 @@ FROM alpine:3.21
 
 RUN apk update && \
     apk add py3-psycopg2 \
+            py3-requests \
             py3-yaml \
             tini
 

tests/docker-compose.yml

@@ -23,7 +23,7 @@ services:
       test: [ "CMD-SHELL", "pg_controldata /var/lib/postgresql/data/ | grep -q 'in production'" ]
       interval: 5s
       timeout: 2s
-      retries: 20
+      retries: 40
     command: >
       postgres -c ssl=on
                -c ssl_cert_file='/etc/ssl/certs/ssl-cert-snakeoil.pem'

tests/run-tests.sh

@@ -6,12 +6,15 @@ versions=( $@ )
 # If we weren't given any versions, test them all
 if [ ${#versions[@]} -eq 0 ]
 then
-    versions=( 9.2 9.6 10 11 12 13 14 15 16 17 )
+    versions=( 9.2 9.4 9.6 10 11 12 13 14 15 16 17 )
 fi
 
 # Image tags to use
 declare -A images=()
 images["9.2"]='9.2'
+images["9.3"]='9.3'
+images["9.4"]='9.4'
+images["9.5"]='9.5'
 images["9.6"]='9.6-bullseye'
 images["10"]='10-bullseye'
 images["11"]='11-bookworm'

zabbix_templates/coverage.py

@@ -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))))