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.
- Fixes the following vulnerabilities (@ptodev):
- Updated OpenTelemetry to v0.102.1. (@mattdurham)
- Components
otelcol.receiver.otlp
,otelcol.receiver.zipkin
,otelcol.extension.jaeger_remote_sampling
, andotelcol.receiver.jaeger
settingmax_request_body_size
default changed from unlimited size to20MiB
. This is due to CVE-2024-36129.
- Components
-
Update Public preview
remotecfg
to usealloy-remote-config
instead ofagent-remote-config
. The API has been updated to use the termcollector
overagent
. (@erikbaranowski) -
Component
otelcol.receiver.vcenter
removedvcenter.host.network.packet.errors
,vcenter.host.network.packet.count
, andvcenter.vm.network.packet.count
.vcenter.host.network.packet.errors
replaced byvcenter.host.network.packet.error.rate
.vcenter.host.network.packet.count
replaced byvcenter.host.network.packet.rate
.vcenter.vm.network.packet.count
replaced byvcenter.vm.network.packet.rate
.
-
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)
-
(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 toprometheus.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) -
Improved filesystem error handling when working with
loki.source.file
andlocal.file_match
, which removes some false-positive error log messages on Windows (@thampiotr) -
Updates
processor/probabilistic_sampler
to use newFailedClosed
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.
- The resource attribute
-
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 theremotecfg
block. (@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 theprometheus.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 atags
config argument to theazure
detection mechanism. It exposes regex-matched Azure resource tags as OpenTelemetry resource attributes.
-
A new
snmp_context
configuration argument forprometheus.exporter.snmp
which overrides thecontext_name
parameter in the SNMP configuration file. (@ptodev) -
Add extra configuration options for
beyla.ebpf
to select Kubernetes objects to monitor. (@marctc)
-
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)
-
pyroscope.ebpf
,pyroscope.java
,pyroscope.scrape
,pyroscope.write
anddiscovery.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'senable_protobuf_negotiation
argument is now deprecated and will be removed in a future major release. Usescrape_protocols
instead and refer toprometheus.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)
-
Fix panic when component ID contains
/
inotelcomponent.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.
- Fixes issue where returned
-
Bubble up SSH key conversion error to facilitate failed
import.git
. (@hainenber)
-
(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 newaws_cloud_map
resolver. (@ptodev) -
Introduce a
otelcol.receiver.file_stats
component from the upstream OpenTelemetryfilestatsreceiver
component. (@rfratto)
-
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 onmetadata_refresh_interval
instead ofprune_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 themimir.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)
-
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
'sextra_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
-
Update
alloy-mixin
to use more specific alert group names (for example,alloy_clustering
instead ofclustering
) to avoid collision with installs ofagent-flow-mixin
. (@rfratto) -
Upgrade Beyla from v1.4.1 to v1.5.1. (@marctc)
-
Add a description to Alloy DEB and RPM packages. (@rfratto)
-
Allow
pyroscope.scrape
to scrapealloy.internal:12345
. (@hainenber) -
The latest Windows Docker image is now pushed as
nanoserver-1809
instead oflatest-nanoserver-1809
. The old tag will no longer be updated, and will be removed in a future release. (@rfratto) -
The log level of
finished node evaluation
log lines has been decreased to 'debug'. (@tpaschalis) -
Update post-installation scripts for DEB/RPM packages to ensure
/var/lib/alloy
exists before configuring its permissions and ownership. (@rfratto) -
Remove setcap for
cap_net_bind_service
to allow alloy to run in restricted environments. Modern container runtimes allow binding to unprivileged ports as non-root. (@BlackDex) -
Upgrading from OpenTelemetry v0.96.0 to v0.99.0.
otelcol.processor.batch
: Prevent starting unnecessary goroutines. open-telemetry/opentelemetry-collector#9739otelcol.exporter.otlp
: Checks for port in the config validation for the otlpexporter. open-telemetry/opentelemetry-collector#9505otelcol.receiver.otlp
: Fix bug where the otlp receiver did not properly respond with a retryable error code when possible for http. open-telemetry/opentelemetry-collector#9357otelcol.receiver.vcenter
: Fixed the resource attribute model to more accurately support multi-cluster deployments. open-telemetry/opentelemetry-collector-contrib#30879 For more information on impacts please refer to: open-telemetry/opentelemetry-collector-contrib#31113 The main impact is thatvcenter.resource_pool.name
,vcenter.resource_pool.inventory_path
, andvcenter.cluster.name
are reported with more accuracy on VM metrics.otelcol.receiver.vcenter
: Remove thevcenter.cluster.name
resource attribute from Host resources if the Host is standalone (no cluster). open-telemetry/opentelemetry-collector-contrib#32548otelcol.receiver.vcenter
: Changes process for collecting VMs & VM perf metrics to be more efficient (one call now for all VMs). open-telemetry/opentelemetry-collector-contrib#31837otelcol.connector.servicegraph
: Added a newdatabase_name_attribute
config argument to allow users to specify a custom attribute name for identifying the database name in span attributes. open-telemetry/opentelemetry-collector-contrib#30726otelcol.connector.servicegraph
: Fix 'failed to find dimensions for key' error from race condition in metrics cleanup. open-telemetry/opentelemetry-collector-contrib#31701otelcol.connector.spanmetrics
: Addmetrics_expiration
option to enable expiration of metrics if spans are not received within a certain time frame. By default, the expiration is disabled (set to 0). open-telemetry/opentelemetry-collector-contrib#30559otelcol.connector.spanmetrics
: Change default value ofmetrics_flush_interval
from 15s to 60s. open-telemetry/opentelemetry-collector-contrib#31776otelcol.connector.spanmetrics
: Discard counter span metric exemplars after each flush interval to avoid unbounded memory growth. This aligns exemplar discarding for counter span metrics with the existing logic for histogram span metrics. open-telemetry/opentelemetry-collector-contrib#31683otelcol.exporter.loadbalancing
: Fix panic when a sub-exporter is shut down while still handling requests. open-telemetry/opentelemetry-collector-contrib#31410otelcol.exporter.loadbalancing
: Fix memory leaks on shutdown. open-telemetry/opentelemetry-collector-contrib#31050otelcol.exporter.loadbalancing
: Support the timeout period of k8s resolver list watch can be configured. open-telemetry/opentelemetry-collector-contrib#31757otelcol.processor.transform
: Change metric unit for metrics extracted withextract_count_metric()
to be the default unit (1
). open-telemetry/opentelemetry-collector-contrib#31575otelcol.receiver.opencensus
: Refactor the receiver to pass lifecycle tests and avoid leaking gRPC connections. open-telemetry/opentelemetry-collector-contrib#31643otelcol.extension.jaeger_remote_sampling
: Fix leaking goroutine on shutdown. open-telemetry/opentelemetry-collector-contrib#31157otelcol.receiver.kafka
: Fix panic on shutdown. open-telemetry/opentelemetry-collector-contrib#31926otelcol.processor.resourcedetection
: Only attempt to detect Kubernetes node resource attributes when they're enabled. open-telemetry/opentelemetry-collector-contrib#31941otelcol.processor.resourcedetection
: Fix memory leak on AKS. open-telemetry/opentelemetry-collector-contrib#32574otelcol.processor.resourcedetection
: Update to ec2 scraper so that core attributes are not dropped if describeTags returns an error (likely due to permissions). open-telemetry/opentelemetry-collector-contrib#30672
-
Use Go 1.22.3 for builds. (@kminehart)
-
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.