Heroku Postgres Metrics Logs

Last updated May 06, 2021

Heroku Postgres Standard and Premium Tier database users will see database-related events on their app’s log stream. This can be useful for recording and analyzing usage over time.

Heroku Postgres Metrics which appear via heroku-postgres are separate from standard alerts emitted from Postgres itself which appear for all applications via postgres.

Log format

Nov 09 23:56:02 source=HEROKU_POSTGRESQL_SILVER addon=postgresql-devcenter-123456 sample#current_transaction=54017686 sample#db_size=16012956319bytes sample#tables=97 sample#active-connections=7 sample#waiting-connections=0 sample#index-cache-hit-rate=0.9239 sample#table-cache-hit-rate=0.93609 sample#load-avg-1m=0 sample#load-avg-5m=0 sample#load-avg-15m=0 sample#read-iops=0 sample#write-iops=0.13333 sample#tmp-disk-used=33849344 sample#tmp-disk-available=72944943104 sample#memory-total=4044960kB sample#memory-free=46920kB sample#memory-cached=3667532kB sample#memory-postgres=20824kB sample#wal-percentage-used=0.06512959334021144

The following attributes appear in application logs for all standard and premium tier databases.

  • source: The database attachment name that the measurements relate to (e.g. HEROKU_POSTGRESQL_VIOLET).
  • addon: The database addon name that the measurements relate to (e.g. postgres-metric-68904).
  • The log line’s timestamp is the time at which the measurements were taken.

Database metrics

These attributes apply to a particular database. They are the same for single-tenant and multitenant database plans.

  • sample#db_size: The number of bytes contained in the database. This includes all table and index data on disk, including database bloat.
  • sample#tables: The number of tables in the database.
  • sample#active-connections: The number of connections established on the database.
  • sample#waiting-connections: Number of connections waiting on a lock to be acquired. If many connections are waiting, this can be a sign of mishandled database concurrency.
  • sample#current_transaction: The current transaction ID, which can be used to track writes over time.
  • sample#index-cache-hit-rate: Ratio of index lookups served from shared buffer cache, rounded to five decimal points. Heroku recommends a value of 0.99 or greater if possible. If your index hit rate is consistently less than 0.99, you should investigate your Expensive Queries or you may need to upgrade your database plan for more RAM.
  • sample#table-cache-hit-rate: Ratio of table lookups served from shared buffer cache, rounded to five decimal points. Heroku recommends a value of 0.99 or greater if possible. If your table hit rate is consistently less than 0.99, you may need to upgrade your database plan for more RAM.
  • sample#memory-postgres: Approximate amount of memory used by your database’s Postgres processes in kB. This includes shared buffer cache as well as memory for each connection.
  • sample#follower-lag-commits: Replication lag, measured as the number of commits that this follower is behind its leader. Replication is asynchronous so a number greater than zero may not indicate an issue, however an increasing value deserves investigation. Read more in Monitoring Followers. This metric is only published for follower databases.

Server metrics

These metrics come directly from the server operating system. For multi-tenant plans, the metrics include other databases running on the shared server, and may be misleading when diagnosing database performance.

  • sample#load-avg-1m, sample#load-avg-5m and sample#load-avg-15m: The average system load over a period of 1 minute, 5 minutes and 15 minutes, divided by the number of available CPUs. A load-avg of 1.0 indicates that, on average, processes were requesting CPU resources for 100% of the timespan. This number includes I/O wait. For databases that have burstable performance, a baseline load average is guaranteed. For more information see the burstable performance section in the technical characteristic article.
  • sample#read-iops and sample#write-iops: Number of read or write operations in I/O sizes of 16KB blocks.
  • sample#memory-total: Total amount of server memory available.
  • sample#memory-free: Amount of free memory available in kB.
  • sample#memory-cached: Amount of memory being used by the OS for page cache, in kB.
  • sample#memory-postgres: Amount of memory being used by Postgres in kB.
  • sample#tmp-disk-used, sample#tmp-disk-available: Amount of bytes used/available on tmp mount.
  • sample#wal-percentage-used: Percentage of the WAL disk that has been used, between 0.0 and 1.0. See this article for more details.

The operating system is designed to maximize the amount of memory it utilizes for its page cache, but can sometimes release that memory if an application requests it. For an estimate of the total amount of memory available to your server, add sample#memory-free and sample#memory-cached together.

PgBouncer Metrics

These metrics are included for any Heroku Postgres server that has a PgBouncer pooler attachment. These metrics are a subset of those viewable by running the SHOW POOLS; command when connected to PgBouncer’s special internal database.

  • sample#client_active: The number of client connections to the pooler that have an active server connection assignment.
  • sample#client_waiting: The number of client connections to the pooler that are waiting for a server connection assignment.
  • sample#server_active: The number of server connections that are currently assigned to a client connection.
  • sample#server_idle: The number of server connections that are not currently assigned to a client connection.
  • sample#max_wait: The longest wait time of any client currently waiting for a server connection assignment.
  • sample#avg_query: The average query time of all queries executed through through poolec connections.
  • sample#avg_recv: The average amount of bytes received from clients per second.
  • sample#avg_sent: The average amount of bytes sent to clients per second.

Feedback

Log in to submit feedback.