History.md

0.4.1 (2015-04-16)

• Fix: node-statsd dependency added to production dependencies

0.4.0 (2015-04-14)

• New: increment statsd on start of master/worker processes

0.3.1 (2015-03-22)

• Small tweak to Graceful._finalLog - instead of relying on winston callback or a setTimeout(fn, 250) to ensure that the final log entry hits the disk, we just do a setTimeout(fn, 0) to give it a chance. And the tests no longer check for that last entry, because it's not reliable. May introduce a feature in the future where the process is allowed to die naturally, since we've already stopped the server, etc. This would require that we and the overall client program unref() all timers.
• Small tweak to Master._restartWorker - it seems that sometimes we would get a disconnect event before the worker had been removed from cluster.workers so we sometimes didn't log this very important error 'No workers currently running!' Now we check our own list at this._workers.
• Overhaul of tests due to this breaking change in node 0.12/iojs. Tests were previously assuming that a request immediately after worker crash would hit the next worker; now the connection is refused until the new worker is up.
• Travis now runs on node 0.12 and iojs 1.4/1.5/1.6 only. Didn't feel like making the tests work on 0.10 as well as the new systems.
• Update dev dependencies

0.3.0 (2014-10-26)

• DomainMiddleware class renamed to GracefulExpress
• All public APIs throw on incorrect parameters
• Graceful is resilient to errors thrown by provided 'check' functions or registered 'shutdown' event handlers
• New mode in GracefulExpress: inProcessTest. If true, will not set up a domain for each request, allowing for easier in-process testing with supertest.
• GracefulExpress: tightens up graceful shutdown by:
  • Closing keepalive connections for all in-progress requests and idle sockets
  • Starting a socket reaper to continually look for and close idle sockets
  • Installing a backstop in case issues get through, return Error with statusCode = 503
• Logging: reduce verbosity, move to thehelp-log-shim to leave logging decisions to the process
• Upgrade to 1.x series of thehelp-last-ditch, which doesn't send SMS by default any longer
• Substantial test coverage added
• Remove peer dependencies in favor of real dependencies
• Remove dependency on lodash

0.2.4 (2014-07-31)

• Update to the more bulletproofed 0.3.x-series of thehelp-last-ditch
• Update dev dependencies

0.2.3 (2014-07-16)

• DomainMiddleware no longer warns on 'unfinished' responses; seems that this event isn't fully reliable in the face of all server topologies. We bind to 'finish', 'close' and 'end' events on the response object, but the handler is wrapped with _.once().

0.2.2 (2014-07-14)

• Startup now allows configuration of logs directory - via options.logs on construction or the LOGS environment variable. Defaults to './logs/'

0.2.1 (2014-05-27)

• Pare down what's in the npm package

0.2.0 (2014-05-25)

Breaking:

• GracefulWorker was renamed to Graceful, and now allows others to register for shutdown notification, doesn't call server.close() directly, and no longer acts as middleware.
• Domain-Middleware now needs a server reference, calls server.close(), registers for Graceful shutdown notification, then sends 'connection:close' for all subsequent requests itself
• Master no longer handles shutdown itself, instead delegating to Graceful - if creating manually, be sure to supply a Graceful instance
• Startup now only calls messenger if no Graceful instance can be found

Other updates:

• Update dev dependencies

0.1.1 (2014-05-01)

• Master.stop() now takes a callback
• DomainMiddlware now logs a warning instead of an error when it gets a res.on('close') event
• Dev dependency upgrade: supertest

0.1.0 (2014-04-28)

• Easy startup of cluster application with root startup method
• Fully functional Master, Startup, GracefulWorker and DomainMiddleware classes
• Graceful shutdown of both master and worker processes with proper exit codes, both for top-level and request-specific errors