PHP agent release notesRSS

New Features

Avoid potential memory exhaustion for long running transactions

• The configuration settings newrelic.transaction_tracer.max_segments_cli and newrelic.transaction_tracer.max_segments_web were added. These settings can be used to limit the number of segments the PHP agent records during a CLI transaction and a web transaction respectively.

  newrelic.transaction_tracer.max_segments_cli defaults to 100000 and thus avoids potential memory exhaustion for long running CLI transactions.

  newrelic.transaction_tracer.max_segments_web defaults to 0, meaning that the PHP Agent shall capture all segments during a web transaction.

  For more information, see the documentation about the PHP agent configuration.

Performance improvements

• The PHP agent now creates segments in a more efficient way, which results in improved performance.

Bug fixes

• The Debian init script now uses pidof instead of ps. This solves issues related to starting the daemon with systemctl on Debian. Previously, the daemon would start and immediately stop.

Bug Fixes

• In 9.6.0, custom outbound headers added to curl requests could be silently removed if both newrelic.cross_application_tracer.enabled and newrelic.distributed_tracing_enabled are disabled. This has been fixed.

New Features

Enhanced visibility into curl_multi_exec calls

• Previously, curl_multi_exec requests appeared as one segment with one total time. Now, we expose the individual segments of a curl_multi_exec request that include individual times and host details. This provides greater visibility as to which URLs are being called and improved ability to troubleshoot slow curl calls.

Configurable daemon start timeout

• The PHP Agent has introduced a new configuration newrelic.daemon.start_timeout. Customers may use this to specify a timeout for the agent to wait for the daemon after a daemon was launched by the agent.

  With this timeout set, the agent will not immediately drop a transaction when the launched daemon hasn't acquired a socket yet, but rather grants the daemon time to do so.

  It is recommended to only set this timeout when instrumenting long-lived background tasks, as in case of daemon start problems the agent will block for the given timeout at every transaction start.

Upgrade Notice

• For cross agent conformance, the agent attribute httpResponseCode has been renamed to http.statusCode. In PHP agent release 9.4, we erroneously introduced httpResponseCode as the replacement for response.statusCode.

  The new http.statusCode agent attribute name will align with other agents and enables standardized alerts and INSIGHTS queries based on a common attribute name.

  Attributes are reported with both the new http.statusCode name and the
  legacy httpResponseCode and response.statusCode attribute names.

  Support for legacy attribute names will be removed in future agent versions.

Known issues and workarounds

Potential memory exhaustion for long running transactions

• See description and recommendations under Known issues and workarounds in the PHP 9.0.0.242 release notes.

New Features

• Support for Real Time Streaming

  • Event data is now sent to New Relic every five seconds, instead of every minute. As a result, transaction, error, and custom events will now be available in New Relic and Insights dashboards in near real time. For more information on how to view your events with a five-second refresh, see the real time streaming documentation.
  • Note that the overall limits on how many events can be sent per minute have not changed. Also, span events, metrics, and trace data is unaffected, and will still be sent every minute.

• Laravel 6 is now fully supported by the PHP agent.

Known issues and workarounds

Potential memory exhaustion for long running transactions

• See description and recommendations under Known issues and workarounds in the PHP 9.0.0.242 release notes.

Bug fixes

• A potential segfault when using PHP 7.4 in connection with framework instrumentation has been fixed.

Known issues and workarounds

Potential memory exhaustion for long running transactions

• See description and recommendations under Known issues and workarounds in the PHP 9.0.0.242 release notes.

Known issues and workarounds

Potential memory exhaustion for long running transactions

• See description and recommendations under Known issues and workarounds in the PHP 9.0.0.242 release notes.

New features in 9.4

Added support for PHP 7.4

• The PHP agent can detect libraries and frameworks that were preloaded via the opcache.preload setting.
  • You can disable this feature by using the following configuration setting: newrelic.preload_framework_library_detection = false.

Request URI attribute is now captured

• We now capture the request.uri attribute. The request URI appears in transaction queries in New Relic Insights.
  • You can disable this attribute by using the following configuration setting: newrelic.attributes.exclude = "request.uri".

Upgrade Notices

• For cross agent conformance, the agent attributes request.headers.User-Agent and httpResponseCode are renamed to request.headers.userAgent and response.statusCode. The value response.StatusCode is changed to an integer.
  • Attributes are reported with both the new and the legacy attribute names.
  • Support for legacy attribute names will be removed in future agent versions.

Bug Fixes

• Since 9.0, transaction traces and span events were not created when newrelic_end_transaction was called inside a PHP function. newrelic_end_transaction now creates transaction traces and span events in any case. It reports all traces and span events for segments that weren't ended at the time of its invocation as unknown.

• Since 9.0, Predis calls weren't instrumented when the Predis client was loaded from a path ending in Predis/Client.php. This has been fixed.

• For inbound distributed tracing payloads with invalid or missing values for pr (priority) and/or sa (sampled) the agent used to assign a default priority of -1 and/or a default sampled value of false to the transaction.

  • This has been fixed, the agent now keeps initial priority and sampled values if the respective values in the inbound distributed tracing payload are missing or invalid.

• The daemon used to erroneously send SIGUSR1 signals to its parent process group in case one of the flags --foreground or --watchdog-foreground was given. This has been fixed.

Known issues and workarounds

Potential memory exhaustion for long running transactions

• See description and recommendations under Known issues and workarounds in the PHP 9.0.0.242 release notes.

New features in 9.3

Trace and entity metadata API calls

• A new API function newrelic_is_sampled() has been added. This call returns true if the current transaction is part of a sampled distributed trace.
• A new API function for obtaining linking metadata been added. newrelic_get_linking_metadata(). This call returns an opaque map of key/value pairs that can be used to correlate this application to other data in the New Relic backend.
• A new API function newrelic_get_trace_metadata() has been added. This call returns a collection of metadata used to identify a trace: trace.id, which provides the currently executing trace's identifier; and span.id, which provides the span identifier associated with the currently executing span.

Configurable connection timeout

• The PHP agent has introduced a new configuration newrelic.daemon.app_connect_timeout. Customers may use this to specify a timeout for the agent to wait for a daemon connection.

  With this timeout set, the agent will not immediately drop a transaction when the daemon hasn't connected to the backend yet, but rather grant the daemon time to establish the connection.

  It is recommended to only set this timeout when instrumenting long-lived background tasks, as in case of connection problems the agent will block for the given timeout at every transaction start.

New features in 9.2

More flexibility for container deployments

• The PHP daemon and agent no longer have to reside on the same host and can now communicate over a IPv4 or IPv6 TCP socket. This can be configured via the newrelic.daemon.address setting in the agent and the --address command line option for the daemon.
• When terminating the New Relic PHP daemon via the SIGTERM signal (and/or the SIGINT signal if started with the -f, --foreground flag), the daemon will now send all buffered data to New Relic prior to exiting.
• The PHP daemon has introduced a new configuration --watchdog-foreground. This keeps the daemon watchdog process in the foreground, whereas the --foreground configuration keeps the daemon worker process in the foreground. The new configuration makes it possible to use the daemon in a blocking way, without losing the additional stability provided by the watchdog process.

Upgrade notices

• The PHP agent has introduced a new configuration newrelic.daemon.address which serves as an alias to newrelic.daemon.port. You may use either to specify the location of the New Relic PHP daemon. If both values are set, newrelic.daemon.address takes precedence.

  Similarly, the PHP daemon has introduced a new configuration --address which serves as an alias to --port. Customers may use either to specify the location of the New Relic PHP daemon. If both values are set, --address takes precedence.

• When starting the daemon as an external process, the daemon will now wait for up to three seconds for the listening port to be ready to receive connections before forking into the background. This usually occurs in (much) less than a second, and most users with this configuration will notice no difference in practice.

  The time that the daemon will wait can be controlled by setting the --wait-for-port setting with a duration. This duration may be 0 to prevent any blocking. If the option is omitted, the default value is 3s.

  Note that this is not the default configuration shipped with the PHP agent, and generally is only used in conjunction with the PHP agent configured with newrelic.daemon.dont_launch set to 3.

  Daemons started in foreground mode (with the --foreground flag) are unaffected, and will behave as before.

Bug fixes

• When duplicating database connections to generate explain plans, the agent will no longer make those connections persistent, even if the original connection was persistent.
• The daemon now synchronously handles critical code paths related to harvesting and merging transaction data. This prevents crashes caused by race conditions.
• Previously, the PHP agent was silently ignoring the setting newrelic.daemon.port if the value was outside of the range 1 - 65535. In this case, it used the default value of /tmp/.newrelic.sock. The PHP agent no longer silently ignores these port values; it now logs these errors in php_agent.log.

Known issues and workarounds

• Potential memory exhaustion for long running transactions. See description under Known issues and workarounds in the PHP 9.0.0.242 release notes.

New Features in 9.1

Symfony 4 support added.

• Web transactions that use the Symfony 4 framework will now be automatically named based on the route or controller name.

Addition of the ability to migrate to Configurable Security Policies (CSP) on a per agent basis for accounts already using High-security mode (HSM).

• When both HSM and CSP are enabled for an account, an agent (this version or later) can successfully connect with either high_security: true or the appropriate security_policies_token configured.

Upgrade notices

• Requests handled by PHP-FPM that result in a 404 error because the script does not exist, or a 403 error because PHP-FPM does not have permission to access the script, will now result in a transaction called 404 or 403, respectively, rather than being named after the request URI. This change was made to prevent metric grouping issues, particularly when sites are being probed by potential attackers.

  If you wish to capture the actual request URI for analysis, it can be attached to the transaction event under the request.uri attribute using the following configuration setting: newrelic.transaction_events.attributes.include=request.uri

Bug fixes

• In version 9.0, Guzzle and Predis execution time could be double counted on application overview and transaction charts in APM, as time could be attributed to both PHP execution and the external or datastore time, respectively. This has been fixed, and charts should now revert back to their previous behavior.
• Restarting a transaction from within a Drupal or WordPress hook could result in a segfault. This has been fixed.

Known issues and workarounds

• Potential memory exhaustion for long running transactions. See description under Known issues and workarounds in the PHP 9.0.0.242 release notes.