Yandex Cloud
Search
Contact UsTry it for free
  • Customer Stories
  • Documentation
  • Blog
  • All Services
  • System Status
  • Marketplace
    • Featured
    • Infrastructure & Network
    • Data Platform
    • AI for business
    • Security
    • DevOps tools
    • Serverless
    • Monitoring & Resources
  • All Solutions
    • By industry
    • By use case
    • Economics and Pricing
    • Security
    • Technical Support
    • Start testing with double trial credits
    • Cloud credits to scale your IT product
    • Gateway to Russia
    • Cloud for Startups
    • Center for Technologies and Society
    • Yandex Cloud Partner program
    • Price calculator
    • Pricing plans
  • Customer Stories
  • Documentation
  • Blog
© 2026 Direct Cursus Technology L.L.C.
Monium
  • Getting started
  • Overview
    • Getting started
    • Overview
      • Data conversions
      • Unified Agent
      • Fluent Bit
    • View logs
    • Query language
    • Log metrics
  • Access management
  • Pricing policy
  • Terraform reference
  • Release notes

In this article:

  • Attribute transformations
  • Priorities
  • Known attributes
  • Invalid attributes
  • Supported OTLP fields
  • Scope level
  • Log level
  1. Logs
  2. Data collection
  3. Data conversions

Input data transformation

Written by
Yandex Cloud
Updated at February 24, 2026
  • Attribute transformations
    • Priorities
    • Known attributes
  • Invalid attributes
  • Supported OTLP fields
    • Scope level
    • Log level

Monium currently receives data only in the OpenTelemetry format. See this proto specification for the API and data formats.

As the OpenTelemetry data model is not a full match of the Monium logical data model (e.g., it does not distinguish between labels and meta), a number of rules and transformations are applied.

Attribute transformationsAttribute transformations

Attributes from resources (ResourceLogs.resource.attributes), instrumentation scopes (ScopeLogs.scope.attributes), and log lines (LogRecord.attributes) are merged and transferred into labels (labels) and metadata (meta) based on the following algorithm:

  1. At each level (resource/scope/log), attributes are collected and sorted between labels and meta:
    • Attributes with names starting with labels. go to labels (the labels. prefix will be deleted).
    • Attributes with names starting with meta. go to meta (the meta. prefix will be deleted).
    • All other attributes go to meta, except for the known attributes.
    • Attributes whose keys start with the _ character or contain non-printable characters are discarded. The number of discarded attributes will be shown in a metric.
  2. labels and meta of all levels (resource/scope/log) get merged; the log-level attributes have the highest priority, the scope-level attributes have lower priority, and resource-level attributes have the lowest priority. For more information, see the table of priorities.
  3. Values in int64 or double type attributes remain numeric. Other types are converted to strings:
    • Bool, a string with the true or false value.
    • Bytes are serialized as a base64 string (RFC 4648).
    • Arrays and dictionaries are serialized similarly to JSON.
  4. Attributes from labels that do not comply with the label limitations are moved to the meta level (the existing attribute gets removed).
  5. Attributes from meta that do not comply with the metadata limitations are discarded. The number of discarded attributes will be shown in a metric.

PrioritiesPriorities

Priorities for attributes in Priorities for attributes in labels

Attribute Resource Scope Log
Valid value of the labels.abc label 2 1 0

(0 for the highest priority, 2 for the lowest)

Priorities for attributes in Priorities for attributes in meta

Attribute Resource Scope Log
abc 8 6 4
meta.abc 7 5 3
Invalid value of the labels.abc label 2 1 0

(0 for the highest priority, 8 for the lowest)

Known attributesKnown attributes

Some system attributes are treated as labels without the labels. prefix:

  • project
  • cluster
  • service, service.name
  • host, host.name, hostname
  • service.component
  • service.version
  • deployment.environment
  • dc

Warning

The request_id attribute is also treated as a known label but will be migrated to meta soon.

Warning

The project, cluster, service, and host system attributes are expected at the resource level.

Specifics of transforming reserved labels:

If a non-empty and non-null value is found by the first key in the list (the highest priority one), it is written to the relevant attribute label. The key that was used is removed. Preservation of the remaining keys in the list is not guaranteed.

Label in otel format (attribute key) Transformed into (attribute label)
service/service.name service
host/host.name/hostname host
labels.component/service.component component
labels.version/service.version version
labels.env/deployment.environment env
labels.dc/dc dc

Invalid attributesInvalid attributes

Attributes that do not comply with the limitations will be deleted and will not go to the long-term storage. The rest of the information from the log line will be saved.

Supported OTLP fieldsSupported OTLP fields

Scope levelScope level

Scope-level fields are transformed into labels:

  • The name field, into the scope.name attribute.
  • The version field, into the scope.version attribute.

The resulting attributes have the lowest priority.

Log levelLog level

Trace and span IDsTrace and span IDs

The trace_id and span_id fields are optional. Field values must have the following lengths:

  • trace_id: 16-byte ID.
  • span_id: 8-byte ID.

Valid IDs will be saved in the system in HEX format as trace.id and span.id, respectively.

For more information, see this specification.

In JSON format, the trace_id and span_id visualization depends on the protocol:

Protocol Visualization
HTTP+JSON HEX string
GRPC+JSON Base64 string

Was the article helpful?

Previous
Overview
Next
Unified Agent
© 2026 Direct Cursus Technology L.L.C.