2019.04.0 Release

Docker tag: quay.io/pganalyze/enterprise:v2019.04.0

New features

• General availability of connection tracing, wait event and vacuum monitoring
  • Added wait events graph
  • Redesigned vacuum monitoring, shows currently running vacuums in more detail
  • Add clickable connection graph to server connections page
  • Track TOAST flag for vacuum runs
  • Connections: Show backend type in addition to application name
  • Note that this is disabled for already existing servers - you need to enable "Activity Sampling" on the server settings page for those servers
• Remove 5 server limitation for high-frequency query statistics
  • This is no longer necessary with new collector versions, since they handle multiple configuration entries in parallel
• Add list-focused organization and server overview pages
  • These are now enabled by default when viewing more than 10 databases in the organization overview, or more than 5 databases in the per-server view
• Add API for adding servers
• Add Log Insights to server-wide namespace, in addition to per-database
  • This also hides all log events that are not database-specific from the per-database view, to avoid any confusion which events affect a particular database instead of the whole server
• Gather correct overall size for partitioned tables
• Change query "I/O %" to "% of All I/O", rework query table logic a bit

Bugfixes

• Index Check: Refactor query <=> table associations to be text-based instead
  • Previously we used table IDs for this, which led to all kinds of problems when tables temporarily disappeared (which led to the Index Check failing)
• Postgres roles: Filter our roles from failed logins
  • This can be confusing to the user as any role that had a failed login attempt, even if it doesn't exist, used to show up in the roles list
• Don't process as default member if invitation is present
  • This avoids the duplicate creation of a new user, if the user has actually been invited by email and is accepting their invite during signup
• Don't mark databases created by activity snapshots as legacy default
• Ignore bogus RDS disk statistics
• Server dashboard: Avoid showing "0" text when some data is not available
• Query origin parsing: Fix crash case for ",=" character combination
• Notification worker: Don't notify for issues that got fixed in the meantime
  • This is particularly useful if we are having statistics processing delays, so old notifications for realtime checks might have resolved themselves
• Historic query stats: Skip very old statistics
• Catch unique constraint violations in query stats processing
• Check configuration: Require modify privileges on database
• Show TOAST size on table detail pages
• Enable features check: Ignore v11 settings that are disabled by default

Performance improvements

• Partition log lines and log files by time
• Vacuum Run: Avoid unnecessary sort for retrieving some stats
• ConnectionGraph: Remove unnecessary data being loaded
• Avoid loading table/index issues unnecessarily
• Reduce allocations for query details endpoint, move to dataload module
• Query slowness check:
  • Optimize index usage
  • Run once a day, instead of every 30 minutes
• Partition backend_queries table to avoid DELETE overhead
• Speed up Log Insights by only loading child lines when necessary
• Avoid UPDATEs on user table for JWT token usage in the API
• Delete unused queries older than 60 days

Other changes

• Various security updates in third-party dependencies
• Move Postgres Roles and replication pages to server namespace
• Improve page title formatting/styling in some prominent cases
  • In particular this aims to avoid confusion as we e.g. have "Dashboard" and "Connections" for bot databases and servers.
• Add explicit "Back to Server" and "Back to Organization" in navigation
  • This makes it easier to explain how to get e.g. to organization-wide settings, which currently would require using the dropdown and clicking again on the currently selected organization