docker run -d -p 3000:3000 -p 8080:8080 --name pw2 cybertec/pgwatch2 # InfluxDB version
# OR
docker run -d -p 3000:3000 -p 8080:8080 --name pw2 cybertec/pgwatch2-postgres # Postgres version
or a specific version
docker run -d -p 3000:3000 -p 8080:8080 --name pw2 cybertec/pgwatch2:x.y.z
When migrating existing "config DB" based setups, all previous schema migration diffs with bigger version numbers need to be applied first from the "pgwatch2/sql/config_store/migrations/" (or /etc/pgwatch2/sql/config_store/migrations/ if using ther pre-built packages) folder. Also it is highly recommended to refresh all the metric definitions as they're constantly improved. For that there's also a refresh_metrics_from_github.sh script provided. YAML based setups don't need any extra actions besides refreshing from Git or installing the new RPM / DEB / Tar packages.
Main changes:
- TimescaleDB metric storage fix - honor retention management via --pg-retention-days parameter. Previously workaround was to configure TimescaleDB directly.
- Gatherer fix - honor user input on InfluxDB SSL cert ignoring in all InfluxDB operations.
- Gatherer improvement - support monitoring all PgBouncer pools from a single config entry when "DB name" left empty.
- A new "Sessions overview" dashboards - together with some new 'backends' columns, gives a high-level user activity overview. PG only.
- Gatherer improvement - do not consider rds_superuser on AWS RDS as superuser anymore to get rid of some out-of-the-box errors.
- Metrics - reduce the amount of data pulled in by limiting most metrics to max 300 entries as mostly the "smaller fish" is not looked at.
- Metrics - ignore TimescaleDB sub-partition (chunk) statistics for tables and only aggregating on the topmost level.
- Metrics - add a 'gce' preset tuned for Google GCE managed PostgreSQL engine.
- Metrics - adjust 'azure' preset by removing some metrics that generated errors by default.
- Metrics - add 'sequence_health' metric + an accompanying get_sequences() helper to be able to detect upcoming sequence overflows.
- Metrics - make some 'backend' fields more consistent. AV workers count now also into "total active per DB".
- Metrics - make "stat_statements" 3x more performant. Thanks @bukem.
- Metrics - add 4 avg query / TX characteristic columns to the "backend" metric to complement max values.
- Config DB - adjust the SQL schema to be re-rollout proof.
- Dashboards - fix "Tables top" for Grafana v7.4+ as there was some backwards-incompatible change.
- Dashboards - only use fully qualified object names so that metrics could drop returning schema and table name separately in next version.
- Dashboards - DB overview unprivileged: replace 1h DB growth with "tups fetched vs returned".
- Documentation - a new chapter on monitoring managed cloud instances, plus many other smaller corrections.
- Docker - start using multi-stage builds to compile the gatherer, reducing the image size slightly.
- K8s / Openshift - Helm chart deployment template corrections. Thanks @pmpetit
- Docker - component update to InfluxDB 1.8.5, Grafana 6.7.5, Go 1.16.3.
NB! If upgrading from an older TimescaleDB storage based setup you should additionally re-roll out the following file on the Metrics DB to make pgwatch2 honor the --pg-retention-days parameter.
Main changes:
- TimescaleDB metric storage - cope with v2.0 breaking API changes and improve indexing.
- New gatherer feature - ability to extract "psutil_*" metrics (CPU, disk usage etc) directly from OS. Use "--direct-os-stats" parameter to enable.
- Gatherer improvement - make Graphite storage connections more resilient. @eshkinkot
- Gatherer improvement - add number of total monitored and currently unreachable databases to the stats port output.
- Metrics - introduce a new 'instance_up' metric with values [0|1] to help with SLA type "uptime percentage" calculations.
- Metrics - correct wal_receiver.replay_lag_b datatype to int8. @eshkinkot
- Metrics - 'replication' now supports also cascaded setups as pg_stat_replication has actually data on "cascading primaries".
- Dashboards - on PG "Alert template" dash replace DB based "Available connections" with "Instance connections used %".
- Docker - give Grafana some more time to roll out its DB schema migrations.
- Docker - introduce an explicit supervisord startup script to better time the startup sequence (5s pauses between components).
- Docker - component update to InfluxDB 1.8.4, Grafana 6.7.5, Go 1.5.8.
- Docker - limit Python PIP to pre v21 versions as it was breaking the build.
- K8s / Openshift - more and better Helm chart templates.
Main changes:
- Gatherer fix - avoid unnecessary internal restarts of metrics gathering introduced in v1.8.1. Was causing much bigger metric data volumes than usual.
Main changes:
- Gatherer fix - avoid unnecessary internal restarts of metrics gathering introduced in v1.8.1. Was causing much bigger metric volumes than usual.
- Metrics fix - remove "spill_bytes" from v13 "replication" metric as it didn't make it into the final Postgres release.
- New feature - support automatic monitoring of whole Patroni "namespaces" with DB type "patroni-namespace-discovery". NB! Only when "etcd" is used as DCS.
- New dashboard - "Table details time lag" to visually compare table access pattern from two time ranges.
- Gatherer improvement - do not connect to config DB in adhoc mode as it's not really not needed.
- Gatherer improvement - add "totalMetricFetchFailuresCounter" to the Stats / Health-check port output.
- Postgres metrics DB - try per-table dropping of old data partitions first, to reduce the chance of running out of locks.
- Web UI improvement - don't require the InfluxDB driver when using Postgres metrics storage.
- Web UI - "stats-summary" page corrections. Note that the page is only for quick verification that metrics are being gathered and deprecated.
- Metrics DB - correct a typo in the TimescaleDB rollout script.
- Metrics - add new metric + helper 'vmstat'.
- Metric helpers - more secure helpers, removing SECURITY DEFINER where not critically needed
- Dashboards - "DB Overview": replace "CPU load" which is not available with tup_fetched vs tup_returned ratio.
- Docker component update - Go 1.15.6.
This release added 1 SQL diff to be applied for Config DB based setups: https://github.com/cybertec-postgresql/pgwatch2/blob/master/pgwatch2/sql/config_store/migrations/v1.8.2-1_patroni_namespace_discovery.sql
Main changes:
- New feature - optional "standby" configs. Enables for example to save on data volumes when having a lot of replicas.
- New PG dashboard - DB overview time lag comparison, enables to visually compare 2 time ranges.
- New PG dashboard - Systems stats time lag comparison, enables to visually compare 2 time ranges.
- Metrics - introduce a new metrics attribute "statement_timeout_seconds" to override per host "statement_timeout".
- Documentation - add separate "docs" section backed by readthedocs: https://pgwatch2.readthedocs.io/.
- Gatherer improvement - allow "adhoc" mode to monitor all DB-s of an instance.
- Gatherer improvement - make startup connection check to config/metricsDB more resilient.
- Gatherer improvement - append instead of overwrite when using JSON-file metrics persistence.
- Metrics - Add an "azure" preset tuned for Microsoft's managed service (@Sascha8a).
- Dashboards - replace "Session count" panel with TPS / QPS for "DB Overview" (PG).
- Dashboards - make "Stat Statements Top" a lot faster for (PG).
- Dashboards - major overhaul of Global DB Overview (PG).
- Dashboards - tag all preset dashboards with "pgwatch2".
- Metric helpers - More secure SECURITY DEFINER helpers, search_path is now inspecting during rollout (@laurenz).
- Docker images - set 500ms wal-fsync-delay for InfluxDB backed images.
- Docker images - add a new "db-bootstrapper" image that can roll out config DB or metric DB schemas.
- Docker component update - Influx 1.8.3, Go 1.15.3.
- Code - switch to Go Modules (@pashagolub).
NB! When migrating existing "config DB" based setups, all previous schema migration diffs with bigger version numbers need to be applied first from the "pgwatch2/sql/config_store/migrations/" (or /etc/pgwatch2/sql/config_store/migrations/ if using ther pre-built packages) folder. Also it is highly recommended to refresh all the metric definitions as they're constantly improved. For that there's also a refresh_metrics_from_github.sh script provided. YAML based setups don't need any extra actions besides refreshing from Git or installing the new RPM / DEB / Tar packages.
This release added 1 SQL diff to be applied for Config DB based setups: https://github.com/cybertec-postgresql/pgwatch2/blob/master/pgwatch2/sql/config_store/migrations/v1.8.1-1_standby_config.sql
- New feature - native support for Pgpool status monitoring + a dash. Makes use of SHOW POOL_NODES and SHOW POOL_PROCESSES commands.
- New feature - allow pausing of metrics collection on certain days/times (with optional time zone support).
- New feature - add support for metric SQL overrides based on versions of installed extensions.
- New feature + metric - privileges (GRANT) change tracking plus also Superuser and "login" role changes + a new panel on "Change events" dash.
- Metrics storage - add support for the TimescaleDB time-series extension with built-in compression.
- Metrics - support for upcoming PostgreSQL v13.
- Metrics - new recommendations engine checks on disabled triggers and ineffective indexing of mostly null columns.
- Metrics - suppress duplicate definition errors in refresh_metrics_from_github.sh "reload" helper.
- Metrics - table_stats / table_io_stats include now also root partition summarized stats.
- Metrics - add a small Python script test_all_metrics.py to validate all metric SQL definitions.
- Metrics - Metrics: add support for monitoring AWS Aurora PG with 'aurora' template.
- Metrics - add an 'rds' preset metrics config tuned for AWS RDS rds_superuser monitoring role usage.
- New dashboard - 'Pgpool Stats' based on newly added Pgpool2 statistics.
- Web UI - update jQuery 3.4.1 to 3.5.1 to plug a security threat.
- Docker images - make deployments more secure by generating a random AES-GCM passphrase for protecting DB passwords if not set by user.
- Docker component update - Influx 1.8.0, Grafana 6.7.4, Go 1.14.4.
NB! When migrating existing "config DB" based setups, all previous schema migration diffs with bigger version numbers need to be applied first from the "pgwatch2/sql/config_store/migrations/" (or /etc/pgwatch2/sql/config_store/migrations/ if using ther pre-built packages) folder. Also it is highly recommended to refresh all the metric definitions as they're constantly improved. For that there's also a refresh_metrics_from_github.sh script provided. YAML based setups don't need any extra actions besides refreshing from Git or installing the new RPM / DEB / Tar packages.
- Gatherer fix - allow using InfluxDB storage without a privileged user, not adjusting any retention policies if retention set to 0.
- Gatherer feature - enable re-routing of metrics, thus slightly different metric definitions can be stored under one metric name.
- Gatherer improvement - "ad-hoc" mode doesn't exit immediately if cannot connect to monitored DB on startup.
- Gatherer improvement - stats port outputs now also counts for metrics reused from cache, datastore failures and average storage write times.
- Gatherer improvement - reduce warnings on detected schema changes to INFO level.
- Gatherer improvement - 'testdatadays' can now be also negative generating historic data.
- New PG dashboard - 'Stat Statements Top (Visual)' based on pg_stat_statements data using new Grafana data links feature.
- Dashboards - Top Tables made faster and added much faster version of Stat Statements Top that assumes no stats reset took place. PG only.
- Dashboards - Recommendation dash (PG): new panel on too frequent checkpoint requests.
- Dashboards - Index overview: don't consider PK indexes as "unused".
- Metrics - new 'stat_statements_no_query_text' metric to hide query texts for "top SQL" statements in a convenient way.
- Metrics - reduce precision for pg_stat_statements based and some others metrics as it took more space and made output harder to read.
- Metrics - add 'logical_subscriptions' metric definition and include to the 'full' preset.
- Web UI - metrics page contents fit better now onto the screen and textareas are resizeable in all directions.
- Web UI - stats-summary page can now show a very high level summary of DB activities also from Postgres metric stores.
- Documentation - many smaller README corrections on custom setup, Web UI configuration, Postgres metrics DB, metrics re-routing.
- Documentation - add a sample regex to use log parsing functionality when using 'stderr' log destination instead of 'csvlog'.
- Documentation - add a K8s template with pgwatch2 as sidecar in Prometheus mode.
- Docker component update - Influx 1.8.0, Grafana 6.7.3, Go 1.14.2.
- Docker - add scripts to build and launch all 'latest' images.
NB! No config DB migrations are needed for this minor release if updating from v1.7.1.
- Gatherer fix - support for PgBouncer v1.12.0+. This latest release changed counter data types.
- Gatherer fix - config DB mode allowed only one continuous discovery Patroni host.
- Gatherer improvement - introduce caching for instance level metrics to reduce load with multi-DB instances. Affects only the 'continous' DB types and is customizable.
- Dashboards - 2 new 'DB Overview' panels + 3 new panels for 'Global health'
- Metrics - new metric + PL/Python helper to fetch WAL-G backup status via the DB.
- Metrics - new metric + PL/Python helper to fetch pgBackRest backup status via the DB.
- Metrics - backend.max_connections, psutil_cpu rounding, other smaller adjustments
- Metrics - add a Bash script to easily refresh SQL based metric definitions (refresh_metrics_from_github.sh).
- Documentation - add info on Prometheus usage, supported PG versions, recommendations for long term setups and exposing logs over the Web UI.
- Documentation - many smaller README corrections, custom installs, backups, Patroni, etc.
- Project structure - get rid of duplicate helper definitions for SQL and YAML modes using symlinks.
- Web UI - metrics page contents fit better now onto the screen and textareas are resizeable in all directions.
- Web UI - metrics page footer now explains supported metric / column attributes
- Web UI - new bulk database management buttons for enabling, disabling, password and config change over all defined DBs
- Packages - include also Web UI sources in DEB/RPM/tar builds so that Web UI could be immediately launched.
- SystemD template - sync paths with those used by DEB/RPM/tar builds
- Docker component update - Influx 1.7.10, Grafana 6.6.2, Go 1.14.
NB! When migrating existing "config DB" based setups, all previous schema migration diffs with bigger version numbers need to be applied first from the "pgwatch2/sql/config_store/migrations/" (or /etc/pgwatch2/sql/config_store/migrations/ if using ther pre-built packages) folder. Also it is recommended to refresh all the metric definitions as they're constantly improved. For that there's also a refresh_metrics_from_github.sh script provided. YAML based setups don't need any extra actions besides refreshing from Git or installing the new RPM / DEB / Tar packages.
- New feature - sever log parsing. Only counts by severity. Assumes local gatherer setup with CSVLOG format by default.
- New feature - "ping" to check connectivity to configured DBs. No metrics gathered. Use the --ping flag / PW2_PING env. var.
- New feature - "realtime" metrics/dashboards, e.g. "Stat activity realtime". Thanks to RealPage Inc. for sponsoring the feature!
- New feature - recommendations engine. List violation of best practices + index advice (pg_qualstats). Thanks to RealPage Inc.!
- New dashboard - "Global health" to summarize uptime / health of all instances, similar to Oracle Enterprise Manager overview.
- Gatherer improvement - when a metric's normal SQL fails, try superuser / direct SQL (if any defined for the metric).
- Gatherer improvement - allow substitution of env variables in YAML monitoring configs. Thanks @rockaut!
- Gatherer improvement - write list of configured DBs back to metrics store for easier health / downtime checks.
- K8s - added HELM chart. Thanks @jinnerbichler!
- Metrics - fix db_stats, tables_stats, table_hashes and bgwriter for old (9.0-9.2) PG versions.
- Metrics - auto-create helper definitions also for Python "psutil" based metrics helpers.
- Metrics - add "get_load_average_copy" helper to get CPU load also without Python (v9.5+).
- Dashboards - add script to import / delete all pgwatch dashboards from command line.
- Dashboards - various minor additions / corrections, especially for Postgres dashboards.
- Dashboards - add last vacuum/analyze infos to "tables top" (PG only).
- Dashboards - add "users" column to "Stat Statements Top".
- Web UI metrics page - make Active DBs listing fit in the column.
- PG metrics DB - save space with partial indexes if there's no "tag" data.
- PG metrics DB - correct Grafana dbname / metric cache table (admin.all_distinct_dbname_metrics) cleanup.
- Docker - make possible to set all verbosity levels via ENV.
- Docker and metrics - make Python 3 the default for helpers / images instead of EOL Python 2.
- Docker component update - Influx 1.7.9, Grafana 6.5.2, Go 1.13.6.
- Infra - add scripts to launch all Postgres versions in Docker and do smoke testing for the pgwatch2 Docker images.
- Infra - RPM / DEB / Tar buils now also include Grafana dashboards + import scripts so all batteries are now included.
- Documentation - many smaller README corrections, custom installs, backups, Patroni, etc.
NB! When migrating old "config DB" based setups, all previous schema migration diffs with bigger version numbers need to be applied first from the "pgwatch2/sql/config_store/migrations/" (or /etc/pgwatch2/sql/config_store/migrations/ if using ther pre-built packages) folder.
- Gatherer improvement - support password/cert authentication for Patroni and etcd
- Gatherer improvement - make pgwatch2 "superuser" aware. Superusers don't need helpers any more for non-Python metrics
- Gatherer fix - in YAML mode statement timeout config file sample didn't match the actual parsing key
- Metrics store fix - correct "metric-dbname-time" model weekly partition creation
- Metrics store improvement - gatherer would not always recover from PG storage failures and restart was needed
- Gatherer improvement - make SystemD service template to re-start on failure
- Gatherer improvement - remove built-in statement timeout override for bloat queries
- Gatherer improvement - always set statement timeout explicitly before any metric queries to avoid a corner case
- Dashboards - new "Postgres Version Overview" dash
- Dashboards - new "Stat Statements SQL Search" dash for finding execution stats for matching SQL texts
- Dashboards - by default filter out pgwatch2 generated metric fetching queries in Stat Statements Top
- Dashboards - Health-check description updates and minor corrections
- Metrics - add 'wal_size' (10+) to 'exhaustive' preset
- Metrics - replace accurate "pgstattuple" based bloat info gathering with SQL based estimates in preset configs
- Metrics - correct older (9.0/9.1) "backends" and "kpi" metrics
- Metrics - add Autovacuum info to "settings" and "table_stats" + display on PG "health-check" dash
- Metrics - "kpi" was failing on replicas for some PG versions
- Docker - fix case where setting PW2_GRAFANASSL=0 still enabled SSL
- Docker - make "Health-check" the default dashboard / splash screen
- Docker component update: Influx 1.7.8, Grafana 6.3.6, Go 1.12.10
NB! When migrating old "config DB" based setups, all previous schema migration diffs with bigger version numbers need to be applied first from the "pgwatch2/sql/config_store/migrations/" (or /etc/pgwatch2/sql/config_store/migrations/ if using ther pre-built packages) folder.
- Config DB fix - allow 'patroni-continuous-discovery' DB type available in YAML mode
- Web UI fix - adding/updating new metrics was broken
- Gatherer fix - correct PG version display, v10.10 was displayed as 10.1
- Gatherer improvement - add --version option to display build Git version
- Metrics DB improvement - metric-dbname-time Postgres storage model using weekly partitions instead of monthly now
- Metrics and metrics DB - add "psql" rollout scripts
- Metrics - add n_live_tup, n_dead_tup to 'table_stats'
- Web UI improvement - replace "psycopg2" dependency with "psycopg2-binary"
- Web UI improvement - don't try to check connection for Patroni dbtype
- Web UI improvement - some sanity checks on adding preset configs
- Dashboards - add unused repl. slots to the Alert Template
- Docker - Dockerfiles not directly dependant on Git now but --build-args + Bash build helper scripts
- Docker component update: Influx 1.7.7, Grafana 6.2.5, Go 1.12.7
- Gatherer feature - add support for Prometheus scraping
- Gatherer feature - Patroni (etcd, Zookeeper, Consul) support (non-password access)
- Gatherer feature - add a flag to monitor DB-s only when they are acting as master / primary
- Gatherer feature - a flag (add-system-identifier / system-identifier-field) to save the "system identifier" with each metric (10+)
- Gatherer fix - handle zeroing of running metric intervals
- Gatherer fix - let the PG driver use .pgpass if no password specified in host settings
- Packaging - Goreleaser support to build DEB, RPM, tarball
- Gatherer improvement - don't start per metric gatherers until connect check OK
- Gatherer improvement -revert persistance maxBatchSize from 5k back to 1k points
- Gatherer improvement - allow partial InfluxDB writes
- Gatherer improvement - support SCRAM-SHA-256 password authentication via Go driver update
- Gatherer improvement - support LibPQ style connection strings in YAML configs
- Gatherer improvement - don't try to auto-create helpers on standbys
- Dashboards - support for Grafana v6 plus minor updates for most dashboards. v5 dashboards will not get updates any more!
- Dashboards - "Health-check" overhaul. PG ver, uptime, transaction wraparound, longest autovacuum and other infos
- Dashboards - a template for alerting with preset thresholds for couple of most important metrics. For PG backend only.
- Dashboards - links going out from "Top stms/table/sproc" dashboards now preserve the selected timerange
- Metrics improvement - skip index_stats gathering for locked indexes
- Metrics - support upcoming PG v12
- Metrics - new "settings" metric based on pg_settings + according panel to "Change events"
- Metrics - capture "server restarted" events when "db_stats" metric enabled (visualized on "Change events" dash)
- Metrics - increase intervals for index/table stats
- Metrics - add a sample bash script to push arbitrary metrics to the pgwatch2 metrics DB externally
- Metrics - add "system identifier" to "wal" to enable auto-grouping of cluster members
- Metrics PG storage - support very long database names also in metric-dbname-time mode. Thanks @henriavelabarbe!
- Config store - add a schema versioning table so that next version schema change diffs could be auto-applied
- Web UI - external component update + according HTML / CSS changes: Bootstrap 3 -> 4, jQuery 3.1 -> 3.4
- Docker components update - Grafana 6.2.4, Go 1.12.6, Influx 1.7.6
NB! When migrating old "config DB" based setups, all previous schema migration diffs with bigger version numbers need to be applied first from the "pgwatch2/sql/config_store/migrations/" folder.
- Gatherer fix - 'continous discovery' worked only in YAML mode
- Metrics fix - pre 9.2 'backends' SQL was incorrect
- Gatherer improvement - 'auto-create helpers' (if "checked") applied to all found DBs now
- Gatherer improvement - "funny" DB names (spaces etc) can be now monitored also with Postgres as storage DB
- Gatherer improvement - issue a warning instead of crash when AES encrypted password does not match required format
- Gatherer improvement - add ability to generate encrypted passwords e.g. for YAML usage, via --aes-gcm-password-to-encrypt param
- Web UI improvement - allow changing password type from plain to AES and vice-versa without re-entering password
- Metrics change - restrict metric names to alphanumerics and underscores. All built-in metrics were already obeying that practice
- Metrics improvement - add a 'create extension if not exists pg_stat_statements' to the get_stat_statements helper SQL
- Docker image with Postgres metrics storage - possible to enable also the "metric-dbname-time" storage model via PW2_PG_SCHEMA_TYPE env. var
- Readme - explanations to available "DB type" options
NB! When migrating old "config DB" based setups, all previous schema migration diffs with bigger version numbers need to be applied first from the "pgwatch2/sql/config_store/migrations/" folder.
- New feature - Support Postgres as metrics storage DB (--pg-schema-type, --pg-metric-store-conn-str, --pg-retention-days)
- New feature - password encryption/decryption with AES-GCM-256 (Gatherer + Web UI)
- New feature - 'verify-ca' and 'verify-full' SSLMODE support (Gatherer + Web UI)
- New feature - Test data generation. New params --testdata-days and --testdata-multiplier (host count)
- New feature - standby-only or master-only metrics (less errors in PG and pgwatch2 logs)
- Gatherer improvement - SystemD support + service file. Thanks @slardiere!
- Gatherer improvement - less error messages when monitored DB or metrics store is down
- Gatherer improvement - disable "connection pooling" by default
- Gatherer improvement - collector logic refactoring, less message passing
- Gatherer improvement - correct "complain only 1x per hour about missing metric definitions"
- Gatherer improvement - new flag for "ad-hoc" mode to control auto-creation of helpers
- New metrics - "wal_receiver" (standby-only), "archiver" (over pg_stat_archiver)
- Metrics improvement - no "public" schema specified in metrics anymore so helpers/extensions can reside anywhere
- Metrics improvement - no explicit grants to "public" for helpers, only to pgwatch2 role
- Metrics - removing DB size from "db_stats" into own "db_size" as it's apparently slow on some FS
- Metrics - new "backup_duration_s" field for "db_stats"
- Metrics - "pg_" removed from "pg_stat_ssl" and "pg_stat_database_conflicts" due to problems with Postgres backend
- Metrics - slight increases to all preset config intervals to be on the conservative side
- Metrics - more efficient "psutil_cpu" helper and useful also with non-persistent sessions
- Dashboard improvement - explicit "agg_interval" variable for most PG overview dashboards as a Grafana workaround
- Web UI improvement - index (default) page set to the "Monitored DBs" page
- Docker - "daemon" image runs now also without any parameters
- Docker - use "tsi1" disk-based index instead of "inmem" for Influx as it's safer for high-cardinality setups
- Docker - new image with Postgres for metrics storage (cybertec/pgwatch2-postgres-storage)
- Docker component update - InfluxDB 1.7.3, Grafana 5.4.3, Go 1.11.4
NB! When migrating old "config DB" based setups, v1.5.0* schema migration diffs need to be applied first from the "pgwatch2/sql/migrations/" folder.
- Metrics fix - KPI for 9.6 corrected
- Gatherer improvement - complain only 1x per hour about missing metric definitions
- Gatherer improvement - re-cycle config DB connections the same way as for monitored DBs
- Gatherer feature - introduce parallel metric fetching per DB, max. 2 queries (hardcoded)
- Gatherer feature - default timeout for built-in bloat metrix set to 30min to run through for 100GB+ tables
- New dash - "Healt-check". Main KPI-s as "single stat" panels with links to according graphs
- New dash - "Index overview". Also supporting "index_stats" metric changes
- New dash - "Tables top". Top-N by size/growth/IUD
- Dashboards - "Replication lag" now has a "dbname" filter like most other dashboards + 2 new panels
- New metric - "archiver" based on pg_stat_archiver to detect WAL shipping issues
- New metric/helper - "wal_size" based on pg_ls_waldir() to detect accumulating WAL folder. PG 10+
- Docker component update - InfluxDB 1.6.4 and Grafana 5.3.2
- Metrics fix - PG10+ 'backends.waiting' column filter was including also non-blocked states
- Dashboard fix - some "Checkpointer/Bgwriter" panels didn't respect "dbname"
- Gatherer feature - allow self-signed InfluxDB SSL certs by default in the gatherer
- Gatherer feature - enable storing gathered metrics to a JSON file with --datastore=json + --json-storage-file=somefile. Allows testing metric gathering and also possibly some data integrations with custom monitoring systems
- Gatherer improvment - better error message on stats interface port collision
- Metrics - casting some fields explicitly to int8
- Metrics - Added XMIN horizon to replication_slots/stat_activity metrics
- Dashboard improvement - increase most aggregation intervals to 5min (was 2m) + some small usability improvements
- Docker component update - Golang 1.11.1, InfluxDB 1.6.3 and Grafana 5.2.4
- Feature - "config file based operation". Now one can run pgwatch2 without the config DB using YAML configs, making automatization easier. See README or help output on --metrics-folder/--config params
- Feature - "Ad hoc" mode. For test/ad hoc purposes the gatherer can now be run from command line, given a JDBC connect string. See README or --adhoc-conn-str param
- Feature - "continous discovery". The gatherer daemon can now periodically scan for new DBs on the cluster and monitor them automatically
- Feature - "custom tags". Now users can add any data (e.g. env. flags/app ID's) to be stored for all metric points as tags in InfluxDB
- Feature - a stats/health interface outputting JSON on internal metric counters for the gatherer. Runs on port 8081 by default, use --internal-stats-port to change
- Feature - "Group" field added for monitored DBs. Enables logical separation and thus running many gatherers on one config DB (sharding)
- Improvement - batching of InfluxDB requests. Huge latency wins over slower connections. Default batching delay is 250ms (changeable)
- Improvement - gatherer daemon now supports < 1s gathering intervals for some extreme use cases
- Improvement - connection pooling on monitored DBs (with 30min recycling). Can be disabled via the --conn-pooling param
- Improvement - set "pgwatch2" as application name on all DB connections for better visibility
- Improvement - InfluxDB retention policy / duration made configurable on command line
- Dashboards - new "DB overview Developer / Unprivileged" dashboard together with an according preset metrics set
- Dashboards - new "System Stats" dashboard together with according helpers / metrics. "psutil" Python package required
- Dashboards - new "Checkpointer/Bgwriter/Block IO Stats" dashboard
- Dashboards - "DB overview" dashboard simplified a bit to be more beginner friendly
- Dashboards - lots of minor corrections e.g. for "Single query details", "Replication"
- Metrics - new metrics "replication_slots", "psutil*" (to take advantage of Python's "psutil" package for OS/system metrics)
- Fix - pg_stat_statements wrapper now compatible with 9.2/9.3
- Fix - make server config change tracking work for PG <9.5
- Fix - correct auto-adding of all DBs in a cluster via Web UI
- Fix - Web UI --pg-require-ssl/--verbose param handling made more robust for env. usage
- Docker - expose the new gather internal statistics port 8081
- Docker - reduce default metrics retention from 90d to 30d
- Docker - Grafana updated to 5.2.2
- Docker - Influx updated to 1.6.1
NB! To migrate from older installations it's also needed to execute v1.4.0 SQL-s from the "pgwatch2/sql/datastore_setup/migrations" folder on the config DB.
- Fix - Openshift/nonroot Docker image was failing on container re-launch due to Postgres SSL false handling
- Fix - setting Grafana SSL via env working now again
- Dashboards - new "Global Overview" dash added for aggregates over all or a set of DBs
- Dashboards - new "Top Sprocs" dash added. Similar to "State Statements Top"
- Dashboards - "Overview" dash "avg. query runtime" more accurate now as pgwatch2 queries are excluded from calculation
- Improvement - stat_statements_calls doesn't need superuser anymore
- Metrics - increase default metric intervals for "multipliable" data i.e. metrics that grow proportional to the amount of objects
- Metrics - added total_time to stat_statements_calls for fast (approximate) avg. query runtime calculation
- Docker - SSL keys are now not re-generated on evey launch
- Web UI - better error handling when Influx DB is not there
- InfluxDB 1.5.1 -> 1.5.3
- Grafana 5.0.4 -> 5.1.3
- Admin UI fix - remove confusing error messages shown on existing connection string updates
- Improvement - stat_statements_calls doesn't need superuser anymore
- Security improvement - only showing pg_stat_activity info on the taget DB
- Metric fix - helper (for superuser auto-create) for bloat was missing
- Gatherer - fix "auto-create" of metric fetching helpers (effective when "Is superuser?" checked on Admin page)
- "Overview" dashboard - minor adjustments
- InfluxDB 1.5.0 -> 1.5.1
- Fix - When using volumes, Postgres Bootstrap was correctly done only for volumes created with "docker volume"
- Fix - Gatherer was always re-logging metric interval change events
- Improvement - 'Stat statements top' dasboard compatibility for older Influx versions (@dankasak)
- Metric improvement - "backends" now has separate parallel query workers counter for PG10+
- Metric/dash improvement - "Sproc details" now based on fully qualified procedure names
- Minor dashboard improvements - "Table details" and "Overview" adjusted for less jumpy graphs on default settings. NB! If migrating an existing setup it is highly recommended to re-import the following dashboards: "Table details", "Overview", "Sproc details"
- Web UI Improvement - showing a warning on "DBs" page if changing the connect string but can't actually connect using it
- README improvements - info on "track_io_timing", component diagram added, new screenshots, project background
- Logging improvement - in some case root cause errors were masked in logs. Better "empty metric SQL" messages
- Logging improvement - remove duplicate event times and milliseconds
- Openshift template - added missing PW2_IRETENTIONDAYS to env vars
- InfluxDB 1.5.0
- Grafana 5.0.4 - old v4 dashboards are now in a separate folder (./grafana_dashboards/v4)
- Go 1.10.1
- Dockerfile/image running as "non-root" user, suitable for example for OpenShift deployments
- Docker VOLUME-s added to Postgres, Grafana, InfluxDB data directories and pgwatch2 persistent config
- Added Dockerfiles for deploying components separately. See the "docker" folder for details
- Grafana security - possible to control anon. access and admin user/passord via env. variables
- New dashboard - AWS CloudWatch overview. One can now easily monitor/alert on on-prem and cloud DBs
- New dashboard and datasource type for PgBouncer stats. Visualizes pgbouncer "SHOW STATS" commands. NB! Requires config DB schema change for existing setups, DB-diff file here
- New dashboard for "Top N" time consuming/frequent/slowest/IO-hungry queries added. Base on pg_stat_statements info. NB! When no SQL info should be leaked, dashboard should be deleted after image start as it shows (parametrized) queries!
- New dashboard - "Biggest relations treemap". Helps to visually detect biggest tables/indexes
- Dashboard chage to "Single query details" - add IO time percentage graph of total time to determine if query is IO or CPU bound. Also showing SQL for the query
- Gatherer daemon - InfluxDB HA support added, i.e. writing metrics to 2 independent DBs. Can be also used for load balancing
- Gatherer daemon - a ringbuffer of max 100k metrics datapoints introduced (i.e. 2GB of RAM) if metric storage is gone. Previously metrics were gather till things blew up
- Gatherer daemon - improved the startup sequence, no load spikes anymore in case of 50+ monitored DBs
- Gatherer daemon - "--iretentiondays" param added to specify InfluxDB retention period (90d default)
- Improved Web UI - nicer errors and providing partial functionality when Postgres or InfluxDB is not available
- Improved Web UI - not showing the "Log out" btn if no authentication was enabled (the default)
- Improved Web UI - new flag ---no-component-logs added to not expose error logs for all the components running in Docker
- Improved Web UI - respecting the --pg-require-ssl param now to force SSL connections to config DB
- README improvements - a new section on custom deployments and some other minor additions
- "Change detection" dashboard/metric improvement - the monitoring DB role is not expected to be superuser anymore
- "Change detection" improvement - showing change event annotations only for the selected DB now
- Improvement - Postgres version for monitored hosts cached for 2 minutes now
- Improvement - Docker image size reduced 20%
- Fix - corrections for "backend" metrics gathering wrapper functions
- Fix for Web UI/Grafana HTTPS mode (outgoing links/logos are now also HTTPS)
- Fix for Docker image Go gatherer - config DB env parameters (PW2_PG*) are now fully respected
- Fix for the "backend" metric - some fields were "null" when using non-superuser. Now there's a wrapper - thanks @jimgolfgti!
- Improvement - table_stats metric skips now over exclusively locked tables
- Improvement - various "Table details" dashboard adjustments (e.g. TOAST/index size info) and according table_stats/index_stats metric adjustments (new full_table_name tag used for searching to counter duplicate table names)
- InfluxDB 1.4.2
- Grafana 4.6.2
- Fix for "panic: runtime error: index out of range" when last table/sproc/index of DB was deleted
- Fix for "max key length exceeded" errors. Stored query texts limited to 16000 chars.
- InfluxDB 1.3.7
- Grafana 4.6.1
- Go 1.9.2
- Fix for "max key length exceeded" errors. Stored query texts limited to 65000 chars.
- InfluxDB 1.3.6
- Grafana 4.5.2
- Deletion of InfluxDB data from the Web UI now possible
- Adding of all databases from a host now possible when leaving "DB name" empty
- All components (Grafana, Postgres, InfluxDB/Graphite) made externally pluggable e.g. you can use your existing InfluxDB to store and access metrics. See README for details
- Fixed login page (no new window popup)
- Not exposing port 8083 anymore as InfluxDB UI was deprecated. Use Chronograf for ad hoc queries from now on
- Better validations and tooltips for the monitored hosts ("/dbs") page in Web UI
- An env. flag not to create the "test" database when launching a pgwatch2 container (-e NOTESTDB=1)
- Make config store DB (Postgres) UTF-8 encoded to avoid problems with non-ascii Grafana dashboard names
- Faster Docker builds when developing, more static parts come first now
- Filtering additionally also on "dbname" besides "queryid" for the "Single query" dashboard
- Corrections and better documentation for the backup/restore script
- InfluxDB 1.3.5 - lots of bugfixes and perf improvements
- Grafana 4.5.1 - query inspection, better query builders and data tables
- Support for Graphite as metric storing database
- SSL support for Grafana and the Web UI
- Support for the upcoming PostgreSQL version 10
- Support for beta/devel versions of Postgres as well as EDB Postgres Advanced Server
- New "change detection" feature and according dashboard
- New stored procedure details dashboard called "Sproc details"
- New approximate table bloat monitoring metric and panel for "DB Overview" dashboard (9.5+)
- New "Queries per Second" panel (QPS label) and according metric stat_statements_calls
- Automatic creation of metric fetching helpers for ssuperusers
- Building the Go daemon with Docker now
- Grafana 4.3.0 - histograms and heatmaps now possible
- Couple of smaller "Overview" dashboard corrections
- InfluxDB update from 1.2.0 to 1.2.2
- Grafana update from 4.1.2 to 4.2.0
- Initial release