Skip to content

Latest commit

 

History

History
622 lines (438 loc) · 33.3 KB

CHANGELOG.md

File metadata and controls

622 lines (438 loc) · 33.3 KB
Raw

Changelog

Contributors should read our contributors guide for instructions on how to update the changelog.

This document contains a historical list of changes between releases. Only changes that impact end-user behavior are listed; changes to documentation or internal API changes are not present.

v1.4.0

Security fixes

  • Add quotes to windows service path to prevent path interception attack. CVE-2024-8975 (@mattdurham)

Breaking changes

  • Some debug metrics for otelcol components have changed. (@thampiotr) For example, otelcol.exporter.otlp's exporter_sent_spans_ratio_total metric is now otelcol_exporter_sent_spans_total.

  • [otelcol.processor.transform] The functions convert_sum_to_gauge and convert_gauge_to_sum must now be used in the metric context rather than in the datapoint context. open-telemetry/opentelemetry-collector-contrib#34567 (@wildum)

  • Upgrade Beyla from 1.7.0 to 1.8.2. A complete list of changes can be found on the Beyla releases page: https://github.com/grafana/beyla/releases. (@wildum) It contains a few breaking changes for the component beyla.ebpf:

    • renamed metric process.cpu.state to cpu.mode
    • renamed metric beyla_build_info to beyla_internal_build_info

Features

  • Added Datadog Exporter community component, enabling exporting of otel-formatted Metrics and traces to Datadog. (@polyrain)
  • (Experimental) Add an otelcol.processor.interval component to aggregate metrics and periodically forward the latest values to the next component in the pipeline.

Enhancements

  • Clustering peer resolution through --cluster.join-addresses flag has been improved with more consistent behaviour, better error handling and added support for A/AAAA DNS records. If necessary, users can temporarily opt out of this new behaviour with the --cluster.use-discovery-v1, but this can only be used as a temporary measure, since this flag will be disabled in future releases. (@thampiotr)

  • Added a new panel to Cluster Overview dashboard to show the number of peers seen by each instance in the cluster. This can help diagnose cluster split brain issues. (@thampiotr)

  • Updated Snowflake exporter with performance improvements for larger environments. Also added a new panel to track deleted tables to the Snowflake mixin. (@Caleb-Hurshman)

  • Add a otelcol.processor.groupbyattrs component to reassociate collected metrics that match specified attributes from opentelemetry. (@kehindesalaam)

  • Update windows_exporter to v0.27.2. (@jkroepke) The smb.enabled_list and smb_client.enabled_list doesn't have any effect anymore. All sub-collectors are enabled by default.

  • Live debugging of loki.process will now also print the timestamp of incoming and outgoing log lines. This is helpful for debugging stage.timestamp. (@ptodev)

  • Add extra validation in beyla.ebpf to avoid panics when network feature is enabled. (@marctc)

  • A new parameter aws_sdk_version_v2 is added for the cloudwatch exporters configuration. It enables the use of aws sdk v2 which has shown to have significant performance benefits. (@kgeckhart, @andriikushch)

  • prometheus.exporter.cloudwatch can now collect metrics from custom namespaces via the custom_namespace block. (@ptodev)

  • Add the label alloy_cluster in the metric alloy_config_hash when the flag cluster.name is set to help differentiate between configs from the same alloy cluster or different alloy clusters. (@wildum)

Bugfixes

  • Fix a bug where the scrape timeout for a Probe resource was not applied, overwriting the scrape interval instead. (@morremeyer, @stefanandres)

  • Fix a bug where custom components don't always get updated when the config is modified in an imported directory. (@ante012)

  • Fixed an issue which caused loss of context data in Faro exception. (@codecapitano)

  • Fixed an issue where providing multiple hostnames or IP addresses via --cluster.join-addresses would only use the first provided value. (@thampiotr)

  • Fixed an issue where providing <hostname>:<port> in --cluster.join-addresses would only resolve with DNS to a single address, instead of using all the available records. (@thampiotr)

  • Fixed an issue where clustering peers resolution via hostname in --cluster.join-addresses resolves to duplicated IP addresses when using SRV records. (@thampiotr)

  • Fixed an issue where the connection_string for the loki.source.azure_event_hubs component was displayed in the UI in plaintext. (@MorrisWitthein)

  • Fix a bug in discovery.* components where old targets would continue to be exported to downstream components. This would only happen if the config for discovery.* is reloaded in such a way that no new targets were discovered. (@ptodev, @thampiotr)

  • Fixed bug in loki.process with sampling stage where all components use same drop_counter_reason. (@captncraig)

  • Fixed an issue (see #1599) where specifying both path and key in the remote.vault path configuration could result in incorrect URLs. The path and key arguments have been separated to allow for clear and accurate specification of Vault secrets. (@PatMis16)

Other

v1.3.1

Bugfixes

  • Changed the cluster startup behaviour, reverting to the previous logic where a failure to resolve cluster join peers results in the node creating its own cluster. This is to facilitate the process of bootstrapping a new cluster following user feedback (@thampiotr)

  • Fix a memory leak which would occur any time loki.process had its configuration reloaded. (@ptodev)

v1.3.0

Breaking changes

  • [otelcol.exporter.otlp,otelcol.exporter.loadbalancing]: Change the default gRPC load balancing strategy. The default value for the balancer_name attribute has changed to round_robin open-telemetry/opentelemetry-collector#10319

Breaking changes to non-GA functionality

  • Update Public preview remotecfg argument from metadata to attributes. (@erikbaranowski)

  • The default value of the argument unmatched in the block routes of the component beyla.ebpf was changed from unset to heuristic (@marctc)

Features

  • Added community components support, enabling community members to implement and maintain components. (@wildum)

  • A new otelcol.exporter.debug component for printing OTel telemetry from other otelcol components to the console. (@BarunKGP)

Enhancements

Bugfixes

  • Fixed a clustering mode issue where a fatal startup failure of the clustering service would exit the service silently, without also exiting the Alloy process. (@thampiotr)

  • Fix a bug which prevented config reloads to work if a Loki metrics stage is in the pipeline. Previously, the reload would fail for loki.process without an error in the logs and the metrics from the metrics stage would get stuck at the same values. (@ptodev)

v1.2.1

Bugfixes

  • Fixed an issue with loki.source.kubernetes_events not starting in large clusters due to short informer sync timeout. (@nrwiersma)

  • Updated ckit to fix an issue with armv7 panic on startup when forming a cluster. (@imavroukakis)

  • Fixed a clustering mode issue where a failure to perform static peers discovery did not result in a fatal failure at startup and could lead to potential split-brain issues. (@thampiotr)

Other

  • Use Go 1.22.5 for builds. (@mattdurham)

v1.2.0

Security fixes

Breaking changes

  • Updated OpenTelemetry to v0.102.1. (@mattdurham)
    • Components otelcol.receiver.otlp,otelcol.receiver.zipkin,otelcol.extension.jaeger_remote_sampling, and otelcol.receiver.jaeger setting max_request_body_size default changed from unlimited size to 20MiB. This is due to CVE-2024-36129.

Breaking changes to non-GA functionality

  • Update Public preview remotecfg to use alloy-remote-config instead of agent-remote-config. The API has been updated to use the term collector over agent. (@erikbaranowski)

  • Component otelcol.receiver.vcenter removed vcenter.host.network.packet.errors, vcenter.host.network.packet.count, and vcenter.vm.network.packet.count.

    • vcenter.host.network.packet.errors replaced by vcenter.host.network.packet.error.rate.
    • vcenter.host.network.packet.count replaced by vcenter.host.network.packet.rate.
    • vcenter.vm.network.packet.count replaced by vcenter.vm.network.packet.rate.

Features

  • Add an otelcol.exporter.kafka component to send OTLP metrics, logs, and traces to Kafka.

  • Added live debugging to the UI. Live debugging streams data as they flow through components for debugging telemetry data. Individual components must be updated to support live debugging. (@wildum)

  • Added live debugging support for prometheus.relabel. (@wildum)

  • (Experimental) Add a otelcol.processor.deltatocumulative component to convert metrics from delta temporality to cumulative by accumulating samples in memory. (@rfratto)

  • (Experimental) Add an otelcol.receiver.datadog component to receive metrics and traces from Datadog. (@carrieedwards, @jesusvazquez, @alexgreenbank, @fedetorres93)

  • Add a prometheus.exporter.catchpoint component to collect metrics from Catchpoint. (@bominrahmani)

  • Add the -t/--test flag to alloy fmt to check if a alloy config file is formatted correctly. (@kavfixnel)

Enhancements

  • (Public preview) Add native histogram support to otelcol.receiver.prometheus. (@wildum)

  • (Public preview) Add metrics to report status of remotecfg service. (@captncraig)

  • Added scrape_protocols option to prometheus.scrape, which allows to control the preferred order of scrape protocols. (@thampiotr)

  • Add support for configuring CPU profile's duration scraped by pyroscope.scrape. (@hainenber)

  • prometheus.exporter.snowflake: Add support for RSA key-pair authentication. (@Caleb-Hurshman)

  • Improved filesystem error handling when working with loki.source.file and local.file_match, which removes some false-positive error log messages on Windows (@thampiotr)

  • Updates processor/probabilistic_sampler to use new FailedClosed field from OTEL release v0.101.0. (@StefanKurek)

  • Updates receiver/vcenter to use new features and bugfixes introduced in OTEL releases v0.100.0 and v0.101.0. Refer to the v0.100.0 and v0.101.0 release notes for more detailed information. Changes that directly affected the configuration are as follows: (@StefanKurek)

    • The resource attribute vcenter.datacenter.name has been added and enabled by default for all resource types.
    • The resource attribute vcenter.virtual_app.inventory_path has been added and enabled by default to differentiate between resource pools and virtual apps.
    • The resource attribute vcenter.virtual_app.name has been added and enabled by default to differentiate between resource pools and virtual apps.
    • The resource attribute vcenter.vm_template.id has been added and enabled by default to differentiate between virtual machines and virtual machine templates.
    • The resource attribute vcenter.vm_template.name has been added and enabled by default to differentiate between virtual machines and virtual machine templates.
    • The metric vcenter.cluster.memory.used has been removed.
    • The metric vcenter.vm.network.packet.drop.rate has been added and enabled by default.
    • The metric vcenter.cluster.vm_template.count has been added and enabled by default.

  • Add yaml_decode to standard library. (@mattdurham, @djcode)

  • Allow override debug metrics level for otelcol.* components. (@hainenber)

  • Add an initial lower limit of 10 seconds for the the poll_frequency argument in the remotecfg block. (@tpaschalis)

  • Add a constant jitter to remotecfg service's polling. (@tpaschalis)

  • Added support for NS records to discovery.dns. (@djcode)

  • Improved clustering use cases for tracking GCP delta metrics in the prometheus.exporter.gcp (@kgeckhart)

  • Add the targets argument to the prometheus.exporter.snmp component to support passing SNMP targets at runtime. (@wildum)

  • Prefix Faro measurement values with value_ to align with the latest Faro cloud receiver updates. (@codecapitano)

  • Add base64_decode to standard library. (@hainenber)

  • Updated OpenTelemetry Contrib to v0.102.0. (@mattdurham)

    • otelcol.processor.resourcedetection: Added a tags config argument to the azure detection mechanism. It exposes regex-matched Azure resource tags as OpenTelemetry resource attributes.

  • A new snmp_context configuration argument for prometheus.exporter.snmp which overrides the context_name parameter in the SNMP configuration file. (@ptodev)

  • Add extra configuration options for beyla.ebpf to select Kubernetes objects to monitor. (@marctc)

Bugfixes

  • Fixed an issue with prometheus.scrape in which targets that move from one cluster instance to another could have a staleness marker inserted and result in a gap in metrics (@thampiotr)

  • Fix panic when import.git is given a revision that does not exist on the remote repo. (@hainenber)

  • Fixed an issue with loki.source.docker where collecting logs from targets configured with multiple networks would result in errors. (@wildum)

  • Fixed an issue where converting OpenTelemetry Collector configs with unused telemetry types resulted in those types being explicitly configured with an empty array in output blocks, rather than them being omitted entirely. (@rfratto)

Other changes

  • pyroscope.ebpf, pyroscope.java, pyroscope.scrape, pyroscope.write and discovery.process components are now GA. (@korniltsev)

  • prometheus.exporter.snmp: Updating SNMP exporter from v0.24.1 to v0.26.0. (@ptodev, @erikbaranowski)

  • prometheus.scrape component's enable_protobuf_negotiation argument is now deprecated and will be removed in a future major release. Use scrape_protocols instead and refer to prometheus.scrape reference documentation for further details. (@thampiotr)

  • Updated Prometheus dependency to v2.51.2 (@thampiotr)

  • Upgrade Beyla from v1.5.1 to v1.6.3. (@marctc)

v1.1.1

Bugfixes

  • Fix panic when component ID contains / in otelcomponent.MustNewType(ID).(@qclaogui)

  • Exit Alloy immediately if the port it runs on is not available. This port can be configured with --server.http.listen-addr or using the default listen address127.0.0.1:12345. (@mattdurham)

  • Fix a panic in loki.source.docker when trying to stop a target that was never started. (@wildum)

  • Fix error on boot when using IPv6 advertise addresses without explicitly specifying a port. (@matthewpi)

  • Fix an issue where having long component labels (>63 chars) on otelcol.auth components lead to a panic. (@tpaschalis)

  • Update prometheus.exporter.snowflake with the latest version of the exporter as of May 28, 2024 (@StefanKurek)

    • Fixes issue where returned NULL values from database could cause unexpected errors.

  • Bubble up SSH key conversion error to facilitate failed import.git. (@hainenber)

v1.1.0

Features

  • (Public preview) Add support for setting GOMEMLIMIT based on cgroup setting. (@mattdurham)

  • (Experimental) A new otelcol.exporter.awss3 component for sending telemetry data to a S3 bucket. (@Imshelledin21)

  • (Public preview) Introduce BoringCrypto Docker images. The BoringCrypto image is tagged with the -boringcrypto suffix and is only available on AMD64 and ARM64 Linux containers. (@rfratto, @mattdurham)

  • (Public preview) Introduce boringcrypto release assets. BoringCrypto builds are publshed for Linux on AMD64 and ARM64 platforms. (@rfratto, @mattdurham)

  • otelcol.exporter.loadbalancing: Add a new aws_cloud_map resolver. (@ptodev)

  • Introduce a otelcol.receiver.file_stats component from the upstream OpenTelemetry filestatsreceiver component. (@rfratto)

Enhancements

  • Update prometheus.exporter.kafka with the following functionalities (@wildum):

    • GSSAPI config
    • enable/disable PA_FX_FAST
    • set a TLS server name
    • show the offset/lag for all consumer group or only the connected ones
    • set the minimum number of topics to monitor
    • enable/disable auto-creation of requested topics if they don't already exist
    • regex to exclude topics / groups
    • added metric kafka_broker_info

  • In prometheus.exporter.kafka, the interpolation table used to compute estimated lag metrics is now pruned on metadata_refresh_interval instead of prune_interval_seconds. (@wildum)

  • Don't restart tailers in loki.source.kubernetes component by above-average time deltas if K8s version is >= 1.29.1 (@hainenber)

  • In mimir.rules.kubernetes, add support for running in a cluster of Alloy instances by electing a single instance as the leader for the mimir.rules.kubernetes component to avoid conflicts when making calls to the Mimir API. (@56quarters)

  • Add the possibility of setting custom labels for the AWS Firehose logs via X-Amz-Firehose-Common-Attributes header. (@andriikushch)

Bugfixes

  • Fixed issue with defaults for Beyla component not being applied correctly. (marctc)

  • Fix an issue on Windows where uninstalling Alloy did not remove it from the Add/Remove programs list. (@rfratto)

  • Fixed issue where text labels displayed outside of component node's boundary. (@hainenber)

  • Fix a bug where a topic was claimed by the wrong consumer type in otelcol.receiver.kafka. (@wildum)

  • Fix an issue where nested import.git config blocks could conflict if they had the same labels. (@wildum)

  • In mimir.rules.kubernetes, fix an issue where unrecoverable errors from the Mimir API were retried. (@56quarters)

  • Fix an issue where faro.receiver's extra_log_labels with empty value don't map existing value in log line. (@hainenber)

  • Fix an issue where prometheus.remote_write only queued data for sending every 15 seconds instead of as soon as data was written to the WAL. (@rfratto)

  • Imported code using slog logging will now not panic and replay correctly when logged before the logging config block is initialized. (@mattdurham)

  • Fix a bug where custom components would not shadow the stdlib. If you have a module whose name conflicts with an stdlib function and if you use this exact function in your config, then you will need to rename your module. (@wildum)

  • Fix an issue where loki.source.docker stops collecting logs after a container restart. (@wildum)

  • Upgrading pyroscope/ebpf from 0.4.6 to 0.4.7 (@korniltsev):

    • detect libc version properly when libc file name is libc-2.31.so and not libc.so.6
    • treat elf files with short build id (8 bytes) properly

Other changes

v1.0.0

Features

  • Support for programmable pipelines using a rich expression-based syntax.

  • Over 130 components for processing, transforming, and exporting telemetry data.

  • Native support for Kubernetes and Prometheus Operator without needing to deploy or learn a separate Kubernetes operator.

  • Support for creating and sharing custom components.

  • Support for forming a cluster of Alloy instances for automatic workload distribution.

  • (Public preview) Support for receiving configuration from a server for centralized configuration management.

  • A built-in UI for visualizing and debugging pipelines.