2018.07.0 Release

Docker tag: quay.io/pganalyze/enterprise:v2018.07.0

New features

• Log Insights 2.0
  • Redesign Log Insights screen for clarity, better zoom-in, links to docs pages
  • Historic Access up to 30 days
    • We've added a new "Jump to Time" selector, as well as other UI elements to allow easy navigation to historic log data, up to 30 days into the past
  • Official support for on-premise systems, Amazon RDS, Amazon Aurora and Heroku Postgres
    • See documentation for details how to set things up
  • Additional log classifications
  • Log Insights Checkpoint Analysis: Add support for 9.3 and 9.4
• Once-per-minute query statistics (instead of every 10 minutes)
  • Accuracy of query data and graphs has been improved, from the previous statistics every 10 minutes, to instead be every minute
  • Graphs update after 10 minutes with the per-minute data (data is loaded in a batch)
• Allow MAILER_FROM environment variable to override password-reset email "From:"
  • Previously the password reset emails were sent from support@pganalyze.com, which causes issues with some SMTP servers that require a specific whitelisted email to be the sender
• Add connections graph on "Connections" page
• Show per-table customized autovacuum settings
• Improved integration with Amazon Aurora
• Add CPU vs I/O time graph on query details page
• Add I/O % to query overview lists
• Add list of "recent issues in last 24 hours" to check config page
• Long running query check: Ignore backups & queries with special comment
• Add "worker" and "snapshot_worker" entry points to support HA configuration
• Add "dbconsole" entrypoint for easier access to the statistics database
• Add summary statistics to vacuum details page (based on log data)
• Add support for ignoring table names for schema processing
  • This currently needs to be set manually on the console

Bugfixes

• Don't store bogus Heroku system statistics for full snapshots
• Handle exclusively locked relations correctly
• Make sequence report work again behind a feature flag
• Upsert: Ensure that all_ids is correct when using DO NOTHING
• Dataload::Queries: Fix bug in fingerprint to query_id mapping
• RetrySnapshot: Adjust code based on snapshots.snapshot_id being available
• Ensure correct index usage when accessing all indices for a database
• Handle "NULL" value (-1) for replication byte_lag correctly
• Normalize UNENCRYPTED PASSWORD statements correctly
• Fix HAVING clause in index slowness check

Performance improvements

• Connection tracing processing speed fixes
  • Use Upsert where possible
  • Avoid N+1 situation in StorageV2::Backends#create_referenced_queries
  • Speed up query that is used to close open backends
• Schema processing speed fixes
  • Speed up schema column/constraint processing using properties_sha1sum
  • Reduce occurrence of schema_table UPDATEs
  • Indices: Reduce number of fields UPSERTed (instead only INSERT)
  • Use structs instead of hashes
  • Only update indices when the table has changed
• Limit maximum number of schema tables processed per server to 5000
  • This limit will be increased in the future, but serves as a good base line for what can be handled currently without issues at the moment
• Add btree_gist indices on backends and vacuum runs
• Add index on schema_indices to support invalid indices check
• Vacuum activity: Preload schema table and postgres role
• Speed up connections list by avoiding N+1 queries
• Add indices to improve loading speed of query details page