Skip to content

Latest commit

 

History

History
755 lines (602 loc) · 31.7 KB

RELEASE_NOTES.md

File metadata and controls

755 lines (602 loc) · 31.7 KB
Raw

Release notes

Worker Pro: easing migration

From time to time Graphile Worker needs to make changes to the database, and these changes might cause pre-existing workers to fail in "interesting" ways. These breaking changes are noted in the release notes below, and typically you need to "scale to zero" to perform these updates - turn off all your existing workers and only then run only new workers.

Worker Pro is a proprietary plugin which, among other things, helps to alleviate this issue. It tracks the running workers and when a new breaking migration is required it has all old workers exit cleanly (once they've finished what they're working on) and no further old workers will be able to start up. Once all old workers have exited, the migration can go ahead as it normally would. This avoids the need to "scale to zero" as workers will communicate with each other to make sure the system as a whole remains consistent.

Read more: Worker Pro Migration.

Pending

  • BREAKING: Jobs and queues are now locked_by their WorkerPool's id rather than the workerId. Be sure to upgrade Worker Pro at the same time if you're using it!
  • New addJobs() JS method to enable efficiently adding a batch of jobs via the JS API
  • DEPRECATION: quickAddJob has been renamed to addJobAdhoc to make it clearer that it's for use in one-off locations (some felt the "quick" referred to the speed it executed, rather than the amount of effort required from the programmer)
  • We'll now warn you if you haven't installed error handlers on the pool, and will only install them ourself if needed
  • Fixes bug where CLI defaults override graphile.config.js settings (by removing CLI defaults)
  • Fix bug where executable tasks had their stdout/stderr ignored; this is now output via logging (thanks @wineTGH).
  • Fix race condition when multiple workers attempt to initialise the database at the same time

v0.16.6

  • Fix bug in workerUtils.cleanup() where queues would not be cleaned up if there existed any job that was not in a queue
  • If errors happen writing to stdout/stderr (e.g. SIGPIPE), we'll trigger a graceful shutdown (and swallow further errors)
  • workerUtils.cleanup({ tasks: ['GC_TASK_IDENTIFIERS'] }) now allows you to specify additional task identifiers to keep (taskIdentifiersToKeep: [...]) in order to reduce impact on other workers
  • graphile-worker --cleanup GC_TASK_IDENTIFIERS will attempt to keep all locally defined task identifiers

v0.16.5

  • Add "timeout" to list of retryable error codes - thanks @psteinroe

v0.16.4

  • Hotfix: remove website dependencies from worker module dependencies

v0.16.3

  • Add "cannot connect now" to list of retryable error codes - thanks @psteinroe

v0.16.2

  • Fix loading tasks on Windows (use URLs rather than file paths) - thanks @hiepxanh
  • Add cleanup function to remove unused queues, stale task identifiers, and permanently failed jobs - thanks @christophemacabiau
  • Fix logger scope for workers - thanks @jcapcik
  • Add helpers.getQueueName() to retrieve the queue name of the currently running job
  • Automatically retry certain internal operations on serialization failure or deadlock detection (useful if you have changed your default_transaction_isolation to serializable or similar)

v0.16.1

  • Fixes issue importing task files that were written in TypeScript ESM format but exported as CommonJS.

v0.16.0

There's a breakdown of these release notes available on the new Worker Website, where we go into more detail about the headline features.

THIS RELEASE INTRODUCES SIGNIFICANT CHANGES, in preparation for moving towards the 1.0 release. Please read these notes carefully.

IMPORTANT: this release is incompatible with previous releases - do not run earlier workers against this releases database schema or Bad Things will happen. You should shut down all workers before migrating to this version, or use Worker Pro. (If you're upgrading from v0.13.0, upgrade to v0.13.1-bridge.0 first and add the Worker Pro plugin to that; deploy it across your fleet, and then proceed to upgrade to v0.16.0.)

General Migration Warnings

  • 🚨 Drops support for node <18
    • As of 24th October 2023, Node 20 is the active LTS and Node 18 is maintainence LTS; previous versions are no longer supported
  • 🚨 Renames all of the tables graphile_worker.* to graphile_worker._private_*
    • We might change these tables in a patch release
    • This has always been the case, these are not a public interface
    • New naming makes it clear that you should not use them and should not rely on their schema being stable
  • 🚨 Removes maxContiguousErrors
    • It wasn't fit for purpose, so best to remove it for now
    • See #307 for more details
  • 🚨 Removes --watch and watch mode in general
    • Now signal handling is improved and with people wanting to use ESM to define modules, it's finally time to remove the experimental watch mode
    • Use node --watch or nodemon or similar instead
    • crontab file is also not watched, so be sure to watch that too!
    • Fixes a lot of weirdness that can happen when you attempt to hot-reload tasks
  • 🚨 Breaking TypeScript changes
    • Lots of any changed to unknown
      • In particular, errors in the event emitter payloads are now unknown rather than any, so you might need to cast
    • Payload is now marked as required in addJob()
      • Set to {} if your task doesn't need a payload

New features

  • Graphile Config and new plugin system
    • Worker is now optionally configurable with graphile-config - configuration can now be read from a graphile.config.ts (or .js, .cjs, etc) file
    • This enables a whole suite of new options and features, including being able to share your preset files across multiple projects!
    • New plugin hooks added - get in touch if you need more!
    • E.g. allows you to replace the task loading code entirely with your own implementation!
  • Support for loading tasks from nested folders
    • (tasks/users/email/verify.js will identify task users/email/verify)
  • Native ESM support
    • Enabled by the plugin system
    • Support for loading both CommonJS and ESM files (including .cjs, .mjs and .js extensions)
  • Compile-to-JS language support
    • Any "compile-to-JS" language can be import()ed
    • Ensure the relevant "loaders" are available
    • e.g. for native TypeScript support you might use NODE_OPTIONS="--loader ts-node/esm" npx graphile-worker
    • List the extensions you support in the configuration file
  • Tasks in non-JS languages (EXPERIMENTAL!)
    • Enabled by the plugin system
    • Any language your shell can execute: python, bash, rust, ...
    • Place an executable file in the tasks/ folder and ensure it's named with the task identifier (extensions ignored)
    • See Loading executable files in the documentation
  • abortSignal: job cancellation (EXPERIMENTAL!)
    • Tasks can honour the abortSignal passed in via helpers to cancel asynchronous work on gracefulShutdown
    • Can reduce waiting for the task to complete during a graceful shutdown; task executor can listen for the abortSignal and decided whether to exit or continue
  • TypeScript typing of tasks
    • New GraphileWorker.Tasks global interface
    • Not recommended, but often requested!
    • Types calls to addJob() and quickAddJob(), and types task executors
    • Read the caveats in the documentation
  • Adds graphile_worker.jobs view
    • A public interface to view details of jobs
    • Stable across patch and minor versions
    • DELIBERATELY excludes the payload field
    • Do not poll this, it will impact performance
    • Do not do expensive filtering/ordering against this, it will impact performance
  • New public force_unlock_workers database function
    • Unlocks all jobs from a list of crashed/terminated worker IDs
  • Crontab: now supports jobKey and jobKeyMode opts (thanks @spiffytech!)
  • Schema
    • Checks that current schema in database isn't more up to date than the current worker
      • Won't be useful until future schema changes
    • Trigger a graceful shutdown if a new Graphile Worker process migrates the database schema
  • Events: add more detail to cron:backfill event
  • Tasks: now use await import(...) rather than require(...), so ESM can be imported
  • Logging: changed format of task completion/failure logs to include attempts/max attempts and to reduce duplicate parenthesis
  • Optimization: Replaces job announcement trigger with calls directly in add_job / add_jobs to reduce queuing overhead

Fixes

  • Fixes graceful shutdown (both manually via .gracefulShutdown() or .forcefulShutdown() and via signal handling)
  • Signals: now releases signal handlers when shut down via the API
  • Fixes bug where queuing 100 jobs in a single statement would only nudge a single inactive worker
    • Now as many workers as necessary and available will be nudged

v0.15.2-bridge.0

TL;DR: if you want to use Worker Pro to ease migration to v0.16.0, upgrade to this version if you're on v0.14.0 or higher, or v0.13.1-bridge.0 if you're on v0.13.0 or lower for Worker Pro support.

This release is a "bridge" release to make migration to v0.16.0 easier. Since v0.16.0 includes breaking database changes, no active workers should be running when the migrations happen, and once the migrations have happened older workers are no longer supported and their usage may lead to weird and undesirable behaviors.

Normally we'd recommend that you "scale to zero" before performing these kinds of migrations, to ensure that no older workers will be running against the DB at the same time; however this release adds support for the (proprietary) Worker Pro plugin which (when used consistently across your entire worker fleet) enables your workers to coordinate, triggering legacy workers to cleanly shut down (and waiting for them to do so) before migrating the database. The Worker Pro plugin also details the intent to upgrade, meaning if new legacy workers start up in the interrim, they will also not start looking for jobs since they know they will be out of date soon. As soon as all running tasks have finished processing (or a configurable timeout has elapsed) the migration will go ahead.

The Worker Pro plugin mentioned above is enabled by the addition of support for graphile-config, the standardized plugin and preset system for the entire Graphile suite. Thanks to this integration, we've been able to add a plugin/hooks system that you can use to customize the behavior of Worker (including implementing some of the behaviors described in the previous paragraph yourself, should you so desire).

This release also adds a startup check that will abort startup if the database already contains breaking migrations that are unsupported by the current worker version, and a significant number of back-ported fixes and new features that didn't require database migrations (including important fixes to the graceful shutdown system).

IMPORTANT: --watch mode is no longer supported. We've removed this from v0.16.0 (see the release notes for that version), you should use node --watch or similar instead. This also removes fauxRequire and all the problems that that had.

IMPORTANT: Node versions before v20 and Postgres versions before v12 are no longer supported. (Node v18 should work, but it segfaults when running the tests which is likely a jest/node --experimental-vm-modules issue which is unlikely affect you at runtime.)

  • Fixes graceful shutdown (both manually via .gracefulShutdown() or .forcefulShutdown() and via signal handling)
  • Removes maxContinguousErrors setting which was poorly implemented and caused more issues than it solved
  • Tracks whether migrations are breaking or not, and:
    • refuses to start if an unsupported breaking migration is present in the database
    • gracefully shuts down if another worker performs a breaking migration
      • NOTE: this is not sufficiently safe, it's just a backstop. If the migration breaks completing or failing of jobs then your worker will be unable to release in-progress tasks even if they're finished; and the worker performing the migrations will not wait for legacy workers to complete. You should shut down your workers before upgrading, or use Worker Pro to handle the situation automatically for you.
  • Adds graphile-config support for presets and plugins
  • Adds more events and hooks
  • runTaskListOnce now uses a WorkerPool internally (to better integrate with the gracefulShutdown logic)
  • Fix WorkerPool.promise to only resolve once everything is handled
  • EXPERIMENTAL; see v0.16.0 for documentation:
    • Adds support for loading tasks from nested folders (e.g. tasks/foo/bar/baz.js will add support for a task with identifier foo/bar/baz)
    • Adds support for turning executable files into tasks (i.e. a task written in python, Rust, or bash)
    • Adds support for loading TypeScript tasks directly (no need to compile to JS, but if you do the JS will have priority)
    • You may see warnings like WARNING: Failed to load task 'README.md' - no supported handlers found for path: '/path/to/tasks/README.md' - you can ignore them (or you can move non-task files out of the tasks folder)
    • Undocumented, experimental and untested preliminary support for cancellable jobs via AbortSignal; upgrade to v0.16.0+ if you want to actually use this
  • A huge number of internal changes

v0.15.1

Fixes issues with graceful worker shutdowns:

  • Deprecates workerPool.release() in favour of (equivalent) workerPool.gracefulShutdown()
  • Fixes workerPool.gracefulShutdown() to shut down gracefully (waiting for jobs to complete)
  • Adds workerPool.forcefulShutdown() to "fail" the running jobs (so they'll be re-attempted elsewhere) and force-release the pool
  • Fixes handling of signals:
    • First termination signal triggers graceful shutdown
    • Signal over next 5 seconds are ignored
    • Second termination signal triggers forceful shutdown
    • Signal over next 5 seconds are ignored
    • Further termination signals are handled by Node (i.e. will likely instantly exit the process)

v0.15.0

Migration files are no longer read from filesystem (via fs module); instead they are stored as strings in JS to enable Graphile Worker to be bundled. The files still exist and will continue to be distributed, so this should not be a breaking change. Thanks to @timelf123 for this feature!

v0.14.0

THIS RELEASE INTRODUCES SIGNIFICANT CHANGES, in preparation for moving towards the 1.0 release. Please read these notes carefully.

IMPORTANT: this release is incompatible with previous releases - do not run earlier workers against this releases database schema or Bad Things will happen. You should shut down all workers before migrating to this version.

IMPORTANT: the initial migration, 000011, in this release cannot run if there are any locked jobs - it will throw a "division by zero" error in this case. Please ensure all existing workers are shut down and any locked jobs released before upgrading to this version.

IMPORTANT: migration 000011 renames the old jobs table, creates a new jobs table with a slightly different format, copies the jobs across, and then deletes the old jobs table. The jobs table itself is not a public interface - you should use the documented SQL functions and TypeScript APIs only - but if you are referencing the jobs table in a database function you may have a bad time.

IMPORTANT: priority, attempts and max_attempts are all now smallint, so please make sure that your values fit into these ranges before starting the migration process. (Really these values should never be larger than about 100 or smaller than about -100 anyway.)

Breaking changes

  • BREAKING: Bump minimum Node version to 14 since 12.x is now end-of-life
  • BREAKING: Bump minimum PG version to 12 for generated always as (expression)
  • BREAKING: jobs.priority, attempts and max_attempts are now int2 rather than int4 (please ensure your values fit in int2 - -32768 <= priority <= +32767)
  • BREAKING: CronItem.pattern has been renamed to CronItem.match
  • BREAKING: database error codes have been removed because we've moved to CHECK constraints

Changes to internals

  • WARNING: the 'jobs' table no longer has queue_name and task_identifier columns; these have been replaced with job_queue_id and task_id which are both ints
  • WARNING: many of the "internal" SQL functions (get_job, fail_job, complete_job) have been moved to JS to allow for dynamic SQL generation for improved performance/flexibility
  • WARNING: most of the triggers have been removed (for performance reasons), so if you are inserting directly into the jobs table (don't do that, it's not a supported interface!) make sure you update your code to be compatible

Features

  • New "batch jobs" feature for merging payloads with a job_key (see README)
  • Significantly improved 'large jobs table' performance (e.g. when a large queue is locked, or there's a lot of jobs queued for task identifiers your worker instance doesn't support, or a lot of failed jobs). Around 20x improvement in this 'worst case' performance for real user workloads.
  • Added new (experimental) much faster add_jobs batch API.
  • Fix error handling of cron issues in 'run' method.
  • CronItem.match can now accept either a pattern string or a matcher function
  • Jobs that were locked more than 4 hours will be reattempted as before, however they are slightly de-prioritised by virtue of having their run_at updated, giving interim jobs a chance to be executed (and lessening the impact of queue stalling through hanging tasks).

v0.13.1-bridge.0

TL;DR: if you want to use Worker Pro to ease migration to v0.16.0, upgrade to this version for Worker Pro support.

This release is a "bridge" release to make migration to v0.14.0 and v0.16.0 easier. Since v0.14.0 and v0.16.0 include breaking database changes, no active workers should be running when the migrations happen, and once the migrations have happened older workers are no longer supported and their usage may lead to weird and undesirable behaviors.

Normally we'd recommend that you "scale to zero" before performing these kinds of migrations, to ensure that no older workers will be running against the DB at the same time; however this release adds support for the (proprietary) Worker Pro plugin which (when used consistently across your entire worker fleet) enables your workers to coordinate, triggering legacy workers to cleanly shut down (and waiting for them to do so) before migrating the database. The Worker Pro plugin also details the intent to upgrade, meaning if new legacy workers start up in the interrim, they will also not start looking for jobs since they know they will be out of date soon. As soon as all running tasks have finished processing (or a configurable timeout has elapsed) the migration will go ahead.

The Worker Pro plugin mentioned above is enabled by the addition of support for graphile-config, the standardized plugin and preset system for the entire Graphile suite. Thanks to this integration, we've been able to add a plugin/hooks system that you can use to customize the behavior of Worker (including implementing some of the behaviors described in the previous paragraph yourself, should you so desire).

This release also adds a startup check that will abort startup if the database already contains breaking migrations that are unsupported by the current worker version, and a significant number of back-ported fixes and new features that didn't require database migrations (including important fixes to the graceful shutdown system).

IMPORTANT: --watch mode is no longer supported. We've removed this from v0.16.0 (see the release notes for that version), you should use node --watch or similar instead. This also removes fauxRequire and all the problems that that had.

IMPORTANT: Node versions before v20 and Postgres versions before v12 are no longer supported. (Node v18 should work, but it segfaults when running the tests which is likely a jest/node --experimental-vm-modules issue which is unlikely affect you at runtime.)

  • Fixes graceful shutdown (both manually via .gracefulShutdown() or .forcefulShutdown() and via signal handling)
  • Removes maxContinguousErrors setting which was poorly implemented and caused more issues than it solved
  • Tracks whether migrations are breaking or not, and:
    • refuses to start if an unsupported breaking migration is present in the database
    • gracefully shuts down if another worker performs a breaking migration
      • NOTE: this is not sufficiently safe, it's just a backstop. If the migration breaks completing or failing of jobs then your worker will be unable to release in-progress tasks even if they're finished; and the worker performing the migrations will not wait for legacy workers to complete. You should shut down your workers before upgrading, or use Worker Pro to handle the situation automatically for you.
  • Adds graphile-config support for presets and plugins
  • Adds more events and hooks
  • Uses JS-ified SQL migrations to help workaround some bundling issues
  • runTaskListOnce now uses a WorkerPool internally (to better integrate with the gracefulShutdown logic)
  • Fix WorkerPool.promise to only resolve once everything is handled
  • EXPERIMENTAL; see v0.16.0 for documentation:
    • Adds support for loading tasks from nested folders (e.g. tasks/foo/bar/baz.js will add support for a task with identifier foo/bar/baz)
    • Adds support for turning executable files into tasks (i.e. a task written in python, Rust, or bash)
    • Adds support for loading TypeScript tasks directly (no need to compile to JS, but if you do the JS will have priority)
    • You may see warnings like WARNING: Failed to load task 'README.md' - no supported handlers found for path: '/path/to/tasks/README.md' - you can ignore them (or you can move non-task files out of the tasks folder)
    • Undocumented, experimental and untested preliminary support for cancellable jobs via AbortSignal; upgrade to v0.16.0+ if you want to actually use this
  • A huge number of internal changes

v0.13.0

  • Remove dependency on pgcrypto database extension (thanks @noinkling)
    • If you have a pre-existing installation and wish to uninstall pgcrypto you will need to do so manually. This can be done by running DROP EXTENSION pgcrypto; after updating to the latest schema.
  • The jobs.queue_name column no longer has a default value (this is only relevant to people inserting into the table directly, which is not recommended - use the add_job helper)

v0.12.2

  • Fix issue when a connect error occurs whilst releasing worker (thanks @countcain)

v0.12.1

  • Jobs with no queue are now released during graceful shutdown (thanks @olexiyb)

v0.12.0

  • Run shutdown actions in reverse order (rather than parallel) - more stable release
  • When an error occurs with the new job listener, reconnection attempts now follow an exponential back-off pattern
  • Allow using Node.js time rather than PostgreSQL time (particularly useful for tests)
  • Refactoring of some cron internals
  • Add noPreparedStatements to the docs

v0.11.4

  • Fixes bug in crontab day-of-week check
  • Exposes parseCronItem helper

v0.11.3

  • Restores Logger export accidentally removed in v0.11.0

v0.11.2

  • Added support for wider range of @types/pg dependency

v0.11.1

  • Handles unexpected errors whilst PostgreSQL client is idle

v0.11.0

  • Export getCronItems so library-mode users can watch the crontab file
  • Replace Logger with new @graphile/logger module

v0.10.0

  • No longer exit on SIGPIPE (Node will swallow this error code)
  • Fix issue with error handling on PostgreSQL restart or pg_terminate_backend
  • Fix a potential unhandled promise rejection

v0.9.0

  • New (experimental) "cron" functionality for regularly scheduled jobs
  • Replace jobs ordering index for improved performance (thanks @ben-pr-p)
    • NOTE: this migration might take a moment if you have a large jobs table
  • New events system lets you monitor what's going on inside Graphile Worker
  • New job_key_mode setting; see README for full details, but summary:
    • defaults to replace (existing behavior, i.e. debouncing)
    • if set to preserve_run_at it will preserve run_at which effectively changes it from debouncing to throttling
    • if set to unsafe_dedupe it will not update the attributes when an existing job with that job key exists, even if that job is already running
  • remove_job now prevents locked jobs from running again on error (and removes their key)
  • Dependency updates

v0.8.1

  • Fix issue with cyclic requires in watch mode

v0.8.0

  • Track revision count for jobs (thanks @lukeramsden)
  • "Forbidden flags" feature for rate limiting (thanks @ben-pr-p)
  • Fix incorrect description of priority - numerically smaller numbers run first (thanks @ben-pr-p)
  • Add support for PG* PostgreSQL envvars

v0.7.2

  • Add --no-prepared-statements flag to allow disabling of prepared statements for pgBouncer compatibility.
  • Fix issue in watch mode where files require()d from inside a task are cached permanently.

(v0.7.0 and v0.7.1 had issues with the experimental watch mode enhancements, so were never upgraded to @latest.)

v0.6.1

  • Official Docker image (thanks @madflow)

v0.6.0

  • Use target es2018 for TypeScript (Node v10 supports everything we need) (thanks @keepitsimple)
  • When task promise is rejected with non-Error, use a fallback (thanks @parker-torii)
  • Support [email protected] and hence Node v14 (thanks @purge)
  • Fix mistake in README
  • General maintenance

v0.5.0

New "Administrative functions", ability to rename graphile_worker schema, and significant overhaul of the codebase in preparation for going to v1.0.

v0.5.0 improvements:

  • Added "Administrative functions" to complete, reschedule or fail jobs in bulk (good for UIs)
  • Added noHandleSignals option to disable our signal handling (if you enable this, make sure you use your own signal handling!)
  • Ability to rename graphile_worker schema
  • Added cosmiconfig for configuration (very few options support this currently)
  • Decrease already negligible chance of worker ID collision (use crypto.randomBytes() rather than Math.random())

v0.5.0 breaking changes:

CLI users: no breaking changes.

Library users: none of the documented (in the README) APIs are affected, except runTaskListOnce and some tiny tweaks to TypeScript types.

The ability to override the SQL schema means that everything in the codebase needs to know this setting. To achieve this:

  • all major APIs now accept options as a configuration parameter
  • where this was optional before it is now required
  • where options was not the first argument, it has been moved to the first argument (for consistency)

As such the following APIs (most of which are internal) have been changed:

  • getTasks(taskPath, watch, logger) -> getTasks(options, taskPath, watch)
  • runTaskList(tasks, pgPool, options?) -> runTaskList(options, tasks, pgPool)
  • runTaskListOnce(tasks, client, options?) -> runTaskList(options, tasks, client)
  • migrate(client) -> migrate(options, client)
  • makeAddJob(withPgClient) -> makeAddJob(options, withPgClient)
  • makeJobHelpers(job, { withPgClient }, baseLogger) -> makeJobHelpers(options, job, { withPgClient, logger? })
  • makeNewWorker(tasks, withPgClient, options, continuous) -> makeNewWorker(options, tasks, withPgClient, continuous)

Also if you're a TypeScript user: we've renamed WorkerSharedOptions to SharedOptions and added a new WorkerSharedOptions. This is particularly relevant if you're using the WorkerUtils class. We've also tweaked what options are available on each of these, but this is unlikely to affect you negatively.

v0.4.0

Performance improvements and ability to efficiently queue jobs from JS.

BREAKING CHANGES:

  • TypeScript:
    • the ID of a job is a string (database bigint), we previously incorrectly stated it was a number.
    • Helpers was renamed to JobHelpers
    • TaskOptions was renamed to TaskSpec
  • queue_name is now nullable (leave it null for maximum parallel performance)
  • when a job is modified using a job_key, the queue_name attribute is now reset like the other attributes

WARNINGS:

  • The database schema has changed; your code should not depend on the database schema, only on the public interfaces (add_job, remove_job, etc), so this shouldn't be an issue.

New features:

  • Significantly enhanced performance
    • Changes database schema such that a job_queue record is only added/checked when necessary
    • Uses prepared statements
    • Can override the PostgreSQL pool size on the CLI (via --max-pool-size)
  • Dedicated API for queueing jobs from JavaScript/TypeScript (makeWorkerUtils / quickAddJob; @mrmurphy, @benjie #60)
  • --once now respects --jobs, so it can run jobs in parallel
  • jobKey is now available via TypeScript API (@tim-field, #78)

Other:

  • Overhauled the perfTest script
  • Upgraded dependencies

v0.3.0-rc.0

v0.3.0-rc.0 was never released as v0.3.0 because we jumped to v0.4.0 too soon.

New features:

  • job_key enables existing jobs to be updated and deleted; can also be used for de-duplication (@gregplaysguitar, @benjie #63)

Fixes:

  • Fixes runner.stop() (@MarkCBall, #66)

v0.2.0

BREAKING CHANGES:

  • The debug task helper has been replaced with a logger helper which is a Logger instance (see README)
  • The -1 shortcut for "run once" never worked; it has been removed
  • Unrecognised command-line arguments will now cause an error to be thrown

New features:

  • Added --schema-only CLI flag for installing/updating the schema (running migrations) only
  • It's now possible to override how logs are output by supplying a logFactory (see README)
  • query helper reduces boilerplate

Fixes:

  • We never needed uuid-ossp so we've removed the requirement (you may want to remove the extension from your DB manually)

v0.1.0

  • Add database 'error' handler to avoid crashes (@madflow #26)
  • DATABASE_URL can now be used in place of connectionString (@madflow, @benjie #20 #27)
  • Improve documentation (@madflow, @archlemon, @benjie #11 #18 #31 #33)
  • Improve testing (@madflow #19 #30)

v0.1.0-alpha.0

Now usable as a library as well as a CLI.

Changes:

  • Renamed a number of internals
    • start -> runTaskList
    • runAllJobs -> runTaskListOnce
    • workerCount -> concurrency
  • Add an easy way to run as a library (run and runOnce methods)
  • CLI code reduced as it uses new library code
  • Implemented linting
  • Exported more methods

v0.0.1-alpha.7

  • Add missing tslib dependency

v0.0.1-alpha.6

  • make poll interval configurable
  • overhaul TypeScript types/interfaces
  • more docs

v0.0.1-alpha.5

  • Fix casting (REQUIRES DB RESET)

v0.0.1-alpha.4

  • add addJob helper

v0.0.1-alpha.3

  • Travis CI
  • Add index.js

v0.0.1-alpha.2

  • Docs

v0.0.1-alpha.1

  • More efficient job trigger
  • Reduce latency

v0.0.1-alpha.0

Initial release.