Node.js agent release notesRSS

Notes:

• Errors that occur in background transactions now have custom parameters copied onto them in the same manner as web transactions.
• Memcached instrumentation has been updated to account for additional arguments that might be passed to the command function that the agent wraps.

Notes

• Custom web transactions can have their names changed by nr.setTransactionName(). Thanks to Matt Lavin for this patch!
• Fixed a bug where Express instrumentation could crash if transaction state was lost in a sub router.

Notes

• Improved the Express instrumentation to be more defensive before doing property lookups, fixing a crash that could happen in an exceptional state.
• Improved logging when the New Relic agent cannot connect to New Relic servers.
• Make Cross Application Tracer header injection less aggressive fixing interaction with other libraries such as riak-js.

Notes

• Added support for Hapi v8.
• briandela contributed a fix for an crash that would occur when using hapi with vhosts.

Fixes

• Fixed a bug that caused some outbound http requests to show up in the New Relic UI as requests to localhost rather than the specified domain.
• The agent no longer reports errors from outbound http requests if they were handled by the user's application

Notes

• The node agent now instruments connections to Oracle Databases using the oracle driver. This patch was contributed by ryanwilliamquinn
• Fixed an issue that would break kraken apps when the node agent was enabled. This patch was contributed by Lenny Markus

Notes

Added support for the the aggregate method on mongodb collections. This patch was contributed by taxilian

Notes

• Fixed a bug in Cross Application Tracing where the agent would sometimes attempt to set a header after headers had already been sent.

• Replaced the logger with one that is handles file writes properly lowering overall resource usage.

  This is a small change with a large impact. fs.createWriteStream returns whether data was queued or not. If it is queued it is recommended to wait on a drain event but this isn't mandatory. Most loggers we've found ignore this event which leads to many writes getting buffered and a rapid increase in native heap size as well as lowering the process's ability to respond to requests.