Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Added
- [Beta] Log forwarding can now collect Syslog logs.
- [Beta] Log forwarding can now collect logs over plain TCP socket connections.
- [Beta] Log forwarding can now be configured via user-provided Fluent-Bit configuration files.
- [Beta] Log forwarding is now able to include custom attributes for each log record as key-value pairs.
For more information, see Forward your logs using New Relic Infrastructure.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Added
- The agent can now retrieve secrets from Vault when running KV Secrets Engine - Version 1.
- Log forwarder entries now include
hostnameas an attribute.
Changed
- Update nri-docker to v1.2.0 (beta Fargate)
Security fixes
- Fixed an issue that could cause secrets to leak into log warning messages when using secrets management.
Bug fixes
- Fixed a log forwarder error that occurred when enabling
eventlogsources. - Fixed an issue that could cause the storage sampler to not return all mount points for a device.
- Avoid spamming the log when agent can't get a process path. This happened mostly when dealing with Windows system-level processes.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Added
Beta: Log forwarding allows you to forward logs to New Relic using the infrastructure agent. For more information, see Forward your logs using the New Relic infrastructure agent.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Added
- The agent now includes the AWS region in the inventory data under metadata/system when running in an AWS VM.
Bug fixes
- Fixed bug on Debian 7 and Centos 5 where the agent would leave child processes running after receiving a service management stop instruction.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Added
- Samplers that are disabled are no longer pre-warmed avoiding errors messages in the log.
Bug fixes
- Fixed an issue with the containerized agent where the agent would crash when the container is stopped
- Fixed an issue in Linux where in some situations the process display name would be empty
- Fixed an issue where in certain situations the unavailability of cpu metrics would lead to a crash
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Added
- Improved auto discovery of ports:
- If a container exposes ports 80 and 443, then
${discovery.port}or${discovery.ports.0}would equal to port80and${discovery.ports.1}to443. As ports are are sorted in ascending order, port8080would have been assigned to number three and so on. - Ports are also tagged by port type (usually TCP):
${discovery.ports.tcp}and${discovery.ports.tcp.0}would equal to80and${discovery.ports.tcp.1}to443.
- If a container exposes ports 80 and 443, then
Bug fixes
- On Windows, the agent can now report processes even when their command line cannot be gathered. The commandLine attribute will be empty in those cases.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Added
- The agent now sends the integration name (
integrationName) and version (integrationVersion) when any events are generated.- For example, when using the Redis monitoring integration
integrationNamewill be set tocom.newrelic.redisandintegrationVersionto1.4.0
- For example, when using the Redis monitoring integration
nri-flexhas been upgraded to versionv1.0.0.- RPM packages generated now include the distribution name in the filename. For example:
- The RHEL8 package
newrelic-infra-<version>-1.x86_64.rpmwill now becomenewrelic-infra-<version>-1.el8.x86_64.rpmwhen published.
- The RHEL8 package
Bug fixes
- The agent no longer panics on startup when reading corrupt caches files.
- When using new configuration format the agent would create inventory as
integration/<binary name>rather thanintegration/<integration name>breaking some charts. For better backwards compatibility the agent will now create inventory using the integration's name.- For example, when using the Redis monitoring integration before it would create inventory under
integration/nri-redisbut with the new update it will create it underinventory/com.newrelic.redis
- For example, when using the Redis monitoring integration before it would create inventory under
- Sample text fields are now truncated to 4095 characters. Before these fields would be rejected and so would never appear in New Relic. Now the fields will appear but will be truncated.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Added
All the distributions of the Agent (packages, containers...) are shipped with
nri-flexandnri-dockerintegrations, so you don't need to install them anymore.- Agent Linux package and containerized agent both now include nri-docker. Before, only the package did.
- Agent Linux package and containerized agent both now include nri-flex. Before, none did.
New containerized agent "bundle" was created from https://github.com/newrelic/infrastructure-bundle , and published at (DockerHub)[https://github.com/newrelic/infrastructure-bundle]. "Bundle" container can now be released and included agent and integration versions could be tracked from the GH repository tags and
versionsfile.
Bug fixes
- Auto-discovery annotations no longer override integration metrics.
- For example: it avoids Kubernetes auto-discovery overriding the cluster_name attributes if it's already set in the integration configuration.
- Removed a memory leak that caused a slow but constant increase in memory and CPU consumption (up to 1%cpu & 5MB per week).
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Bug fixes
- Updated bundled
nri-dockerintegration version to v1.1.1.- This version fixes missing Docker container metrics improving Linux
cgrouppath detection. This issue happened whencgroupwas not mounted in the standard path/sys/fs/cgroup. Now it can discover differentcgrouppaths. - Still cgroup PIDs (process and thread count) is not available on Kernel versions lower than 4.3 see support. Therefore column threadCount won't be available for these systems. For instance Amazon-Linux v1 runs on kernel 4.13.
- This version fixes missing Docker container metrics improving Linux
- Fixed permissions to
/tmp/nri-integrationsdirectory while updating the agent from root mode to privileged or unprivileged mode.- This changes the owner of the
/tmp/nri-integrationsfolder, which is used by some integrations and the Infrastructure integrations SDK to cache data.
- This changes the owner of the
- Fixed issue when running in Kubernetes where an integration configuration change provided by
ConfigMapwouldn't be picked up by the agent, breaking the hot config reload feature for integrations.
Notes
A new version of the agent has been released. Follow standard procedures to update your Infrastructure agent.
Features
- Added IPv6 support during hostname resolution via DNS.
- In verbose mode, more detailed logging during hostname resolution.
Bug fixes
- Fixed a bug from the new integrations' engine added in the version 1.8.0, which caused any JSON line larger than 64 KB to be ignored. This does not affect to the integrations that are configured with the format prior to 1.8.0, which is still supported.
- Fixed
${config.path}replacement in new integrations' configuration file. When an integration configuration contained aconfigorconfig_file_templatesection and the${config.path}variable was used in an environment variable or command-line argument, the correct path was set only during the first execution, remaining the same for the rest of executions, even if the path changes. - Fixed Linux packaging metadata.
- Updated the
Vendorfield from "<jenkins@401c99d99b5f>" to "New Relic, Inc.", fixing the problem that some customers where having on SLES v11.2 Suse, or higher, due to the "repository vendor stickiness", which protects apps from accidentally replacing stable versions of installed packages with experimental versions from a different repository when updating. - Added extra metadata fields:
License,Maintainer,URL,Description,DistributionandSummary
- Updated the