v1.0.0

Notable

• Added support for named prepared statements
• Added support for read replicas and query load balancing
• Added a client_idle_timeout option
• New docs build process and website (Github hosted)
• Docs for migrating from PgBouncer
• auth_query with md5 support
• native pool_mode to proxy direct connections to Postgres
• New metric for unique connected tenants
• Cache metadata database queries for faster tenant info lookups
• Bug fixes

What's Changed

• feat: added a metric for the number of connected tenants by @abc3 in #154
• fix: scram channel binding by @abc3 in #157
• fix: the total count of connected clients for a tenant by @abc3 in #155
• fix: display 0 in the supavisor_connections field after the last disconnection by @abc3 in #158
• feat: add a mode type to the pool id by @abc3 in #160
• chore: publish a tarball to prod by @abc3 in #161
• feat: add client_idle_timeout option by @abc3 in #165
• chore: update GH actions dependencies by @delgado3d in #166
• chore: bump ecto_sql and logflare_logger_backend by @abc3 in #170
• feat: add an option to set the pool's strategy, default is fifo by @abc3 in #173
• feat: support auth_query with md5 by @abc3 in #175
• fix: error message during md5 auth by @abc3 in #177
• fix: cache for user query, server_name_indication, change log level by @abc3 in #179
• fix: add sni in get_secrets by @abc3 in #181
• fix: host in db_handler is already a charlist by @abc3 in #182
• chore: move docs to mkdocs by @chasers in #186
• docs: auto publish docs on main by @chasers in #189
• fix: install mkdocs-material by @chasers in #192
• docs: more docs by @chasers in #190
• feat: add gh workflow by @J0 in #180
• feat: add native mode by @abc3 in #188
• fix: handle http request in the native mode by @abc3 in #193
• feat: handle cancelling requests by @abc3 in #194
• fix: prevent client <-> db locking by @abc3 in #195
• fix: update doc links by @abc3 in #198
• fix: docs set defaults on tenant for example by @chasers in #199
• fix: docs enumerate tenant and user configs with auth_query by @chasers in #201
• feat: add metrics for connection success/failure and improve error handling by @abc3 in #200
• fix: add docs for migrating from pgbouncer by @chasers in #205
• chore: change from native mode to session mode by @abc3 in #210
• feat: handling of named prepared statements by @abc3 #207
• feat: add support for read replicas by @abc3 in #162

New Contributors

• @delgado3d made their first contribution in #166

Full Changelog: v0.9.0...v1.0.0

v0.9.0

Notable

• SSL support upstream and downstream - clients can connect with SSL and SCRAM-SHA-256 authentication
• Optionally enforce SSL connections - set enforce_ssl to true on the tenant to force clients to connect with SSL
• Authentication query support like PgBouncer - define an auth_query on the tenant to authenticate users dynamically
• Metrics endpoint per tenant - /metrics/:external_id will respond with metrics filtered for a single tenant by external_id
• Region and instance id metadata in logs - filter your logs easily by region or instance_id
• Optionally limit client connections per tenant - set default_max_clients on the tenant and optionally max_clients on the user to limit inbound connections per tenant

What's Changed

• docs(openapi): match tenant update implementation by @ChristianAlexander in #65
• New logo 🔥 by @chasers in #71
• Implement multiple pools per tenant by @abc3 in #73
• fix the 'list tenants' endpoint by @abc3 in #84
• Implement session mode functionality by @abc3 in #79
• Implement Websocket proxy by @abc3 in #82
• feat: handle MD5 authentication by @altjohndev in #89
• feat: add version check to CI via VERSION file by @J0 in #78
• Add logflare support by @abc3 in #96
• Systemd service by @abc3 in #95
• Handle http request by @abc3 in #98
• Add prefix to migration AddTimeoutToUsers by @acco in #99
• Add credo config by @abc3 in #108
• Prevent linking to the db process on startup when transaction mode by @abc3 in #107
• Check creds before adding to db and wait timeout for cold start by @abc3 in #101
• Add Postgres strategy for libcluster by @abc3 in #110
• chore: set min/max values for inet_dist_listen by @abc3 in #113
• Add connection pool stop on db credential change by @abc3 in #114
• chore: add prefix to osmon metrics by @abc3 in #117
• chore: support ipv4 and ipv6 for tenant's db by @abc3 in #115
• fix: pg version without platform by @abc3 in #118
• feat: add mix tasks for generating release artifacts by @abc3 in #119
• SSL support for upstream connections by @abc3 in #124
• SSL support for downstream connections by @abc3 in #129
• Option to allow only SSL connection by @abc3 in #130
• feat: auth_query support by @abc3 in #132
• fix: start proxy on two ports by @abc3 in #136
• fix: update the query for retrieving user creds by @abc3 in #137
• fix: use username from the connection in process names by @abc3 in #141
• feat: add region and instance id to logger by @abc3 in #142
• feat: limit client connections by @abc3 in #144
• feat: api endpoint to terminate tenancy tree connection by @abc3 in #145
• fix: query count and network usage metrics by @abc3 in #149
• feat: metric endpoint per tenant by @abc3 in #147

New Contributors

• @kiwicopple made their first contribution in #53
• @MildTomato made their first contribution in #62
• @zorbash made their first contribution in #64
• @ChristianAlexander made their first contribution in #66
• @altjohndev made their first contribution in #89
• @J0 made their first contribution in #78

v0.1.0

• Connect to multiple different Postgres instances/clusters
• Transaction pooling
• Cluster-able
• Resilient during cluster resizing
• Supports rolling and blue/green deployment strategies
• NOT run in a serverless environment
• NOT dependent on Kubernetes
• Easily understand throughput by a tenant, tenant database or individual connection
• Prometheus /metrics endpoint
• OpenAPI spec at /api/openapi
• SwaggarUI at /swaggerui
• When deployed as a Supavisor cluster and a node dies connection pools should be quickly spun up or already available on other nodes when clients reconnect
• Connection buffering
• Brief connection buffering for transparent database restarts or failovers