Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Added
- Optional JSON log format (doc coming soon).
Changed
- Startup info and errors are now logged into the log file since the beginning.
- Log messages prior to configuration being fully loaded.
- Improved error catching and logging on fatal failures.
- Improved log messages. Most of the messages are now static, not inlined with context values, so messages are more easily filterable/greppable, and context fields are provided attached.
Example old message:
Containerized agent found with container id VALUE
Example new message:
Containerized agent found in container containerID: VALUE
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Changed
- Allow enabling http-server (required for Statsd integration) to work in forwarder mode.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Added
Improved logging and troubleshooting features:
- Logging agent config options at startup (with obfuscation of sensitive data).
- Logging all the config options of the loaded integrations at startup (with obfuscation of sensitive data).
- Improve log messages related to integrations.
- Integrations health checks
- New binary called newrelic-infra-ctl for sending notifications to the agent for:
- Enabling verbose logs for 5 minutes
- Integrations health check
Agent daemon notification
For notifying the agent to perform actions, the newrelic-infra-ctl command was created. It has the sole responsibility of sending a notification to the agent process that's running inside the host.
In Linux this communication happens via POSIX signals. newrelic-infra-ctl first tries to find if there is an agent process running. If it finds it, it checks if the process is running inside a Docker container. This is because signals to Docker containers are sent with the docker kill command. Once it knows the pid and the signaling method to use, it sends a SIGUSR1 signal to the agent. The agent just has to set a signal handler for when it receives a SIGUSR1.
In Windows, things are a little different. The only signals that we could send to the process were exit signals. In Windows processes communicate by sending messages to their respective windows, but you might be wondering (as I was), the agent is a daemon process so it doesn't have a window. The solution, easy, a hidden window. On startup the agent creates a hidden window, and sets a handler that listens for messages that arrive in that window. newrelic-infra-ctl finds that windows and sends messages to it.
Changed
- Decorate samples with reportingAgent attribute and replace loopback address if found in reportingEndpoint and reportingEntityKey samples attributes.
- Do not log config.pid_file property in the Windows agent.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Changes
- Minor internal architectural improvements. No impact expected.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Features
- Add support for Red Hat Enterprise Linux 8 and CentOS 8
Improvements
- Improve log messages related to startup permissions errors.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Bug fixes
- When the configuration variable
startup_connection_retries
is set to -1, the Agent will keep checking forever the connectivity at startup to avoidCrashLoopBackOff
states in K8S in case of failures. 'startup_connection_retry_time' has been removed and now the Agent uses an exponential backoff algorithm. - Improve the exponential backoff algorithm the Agent uses when posting metrics fails due to a non 2xx request response.
- Filter duplicated device entries in the
StorageSample
.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Bug fixes
- Fix post installation script for upstart.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Changes
- Removed libcap dependency since it's only used for privileged mode.
Bug fixes
- Lower log level for Selinux errors.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Improvements
- Added HTTPS proxy certificate verification. You can enable it with the
proxy_validate_certificates: true
configuration option. When this option is set totrue
and you setup an HTTPS proxy with self-signed certificates, you may need to specify the certificates location with theca_bundle_file
orca_bundle_dir
configuration options. - Added a Proxy Configuration Plugin that submits non-sensitive proxy configuration information as inventory. To disable it, set the
proxy_config_plugin: false
configuration option. - Warn on RPM plugin package parsing error.
Changes
- Increased the intervals of some samplers
- Increased ProcessSample default interval from 10 to 20
- Increased StorageSample default interval from 5 to 20
- Increased NetworkSample default interval from 5 to 10
- When the configuration option
startup_connection_retries
is set to -1, the Agent, at startup, will keep checking forever the connectivity. - The configuration option
startup_connection_retry_time
has been removed and now the Agent uses an exponential backoff algorithm. - The Windows Agent does not ignore
A:
andB:
drives anymore, as they may be mapped to actual Hard Disks. If those drives are reported because they map into a removable drive and you don't want the agent to report them, you can set thewin_removable_drives: false
configuration option.
Bug fixes
- Fixed a bug that may cause some configuration options to be ignored when settting via environment variables (e.g.
NRIA_IGNORE_RECLAIMABLE
orNRIA_PROXY_VALIDATE_CERTIFICATES
) - Fixed a bug in the Linux agent that prevented the agent from loading the system root certificates when the
ca_bundle_dir
orca_bundle_file
options were specified. This bug can't be fixed in the Windows Agent so the user will need to place in theca_bundle_dir
a copy of the New Relic certificate.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Improvements
- Added replacement of on-host integration's remote entity-names; when a loopback address is found, it will be replaced with agent entity-name. This change will be applied in the entity key and the hostname metric field (if present), when the data comes from an integration using protocol V3. See protocol V3 documentation for further details.
Bug fixes
- Fixed unreported processes issue caused by the inability to parse a different format of /proc//stat.
- Fixed a problem that caused the Windows agent to submit
os:"unknown"
.
Security updates
- Fixed a low security issue that caused the Windows agent to periodically access
C:\etc
folder.