v4.0.5 Retry keys

Fix an issue with the retry plugin in which keys for complex args were being stored as [Object object] in all cases

• by @mklilley via #182

v4.0.4: Bugfix

• fix bug which might cause multiWorker to always work less than minTaskProcessors
  • by @evantahler via 67f0a70

No-Op Error Plugin; Ensure Integers for enqueue delays

• Demonstrate an 'error-ignoring' resque plugin
  • by @gcoonrod via #179
• Ensure that time delays / time stamps are integers for queue.enqueueAt and queue.enqueueIn
  • by @jamesjnadeau via #178

Offload pid parsing to OS to reduce returned buffer size

When getting a list of all runing PIDs on this host (so we can cleanup any previously crashed workers), we had previously taken in the whole PS string and parsed it in Node. If your OS is busy, this list can be quite large and crash your application. Now, we'll offload some of the parsing of this list to AWK (another process) and only get back a list of PIDs to parse.

This shouldn't harm compatibility, as AWK is ubiquitous as GREP.

Fix worker queue reporting on success

This PR changes the internal worker methods of worker.fail and worker.succeed to use (and wait for) a callback internally. This prevents the followup methods, specifically worker.poll to fire before they complete.

This was causing a problem where worker.poll could potentially change the value worker.queue before the success (or failure) event fires, which reads from worker.queue

Remove Domains

From the new README:

It is very important that your jobs handle uncaughtRejections and other errors of this type properly. As of node-resque version 4, we no longer use domains to catch what would otherwise be crash-inducing errors in your jobs. This means that a job which causes your application to crash WILL BE LOST FOREVER. Please use catch() on your promises, handle all of your callbacks, and otherwise write robust node.js applications.

If you choose to use domains, process.onExit, or any other method of "catching" a process crash, you can still move the job node-resque was working on to the redis error queue with worker.fail(error, callback).

---

This is a breaking change which will trigger the v4 release

Solves #166. Visit this issue for a discussion of the possible alternative solutions

Update worker untrack commands to work across cluster (no multi)

(Solves #158).
Removes a multi delete command when untracking workers to prevent errors when using a redis cluster.

remove event listiners from redis when connections end

Soves an issue where stopping and starting multiWorker child workers when passing in a redis connection directly would cause an event-emitter warning.

See #155 for more details.

v3.0.2: Check if setInterval().unref() exists before call

Test if unref() exists before calling it. This allows node-resque to work in a browser context (such as an electron render process).

• by @semiosis via #154

v3.0.1

Speed up the shutdown behavior of mutliWorker