Data processing and pipelines

updated: 2020-06-08 01:26

Data processing and pipelines¶

The mechanism by which data is processed is called a pipeline. Pipelines, at the configuration level, describe a coupling between sources of data and the corresponding sinks for transformation and publication of data. This functionality is handled by the notification agents.

A source is a producer of data: samples or events. In effect, it is a set of notification handlers emitting datapoints for a set of matching meters and event types.

Each source configuration encapsulates name matching and mapping to one or more sinks for publication.

A sink, on the other hand, is a consumer of data, providing logic for the transformation and publication of data emitted from related sources.

In effect, a sink describes a chain of handlers. The chain starts with zero or more transformers and ends with one or more publishers. The first transformer in the chain is passed data from the corresponding source, takes some action such as deriving rate of change, performing unit conversion, or aggregating, before publishing.

Pipeline configuration¶

The notification agent supports two pipelines: one that handles samples and another that handles events. The pipelines can be enabled and disabled by setting pipelines option in the [notifications] section.

The actual configuration of each pipelines is, by default, stored in separate configuration files: pipeline.yaml and event_pipeline.yaml. The location of the configuration files can be set by the pipeline_cfg_file and event_pipeline_cfg_file options listed in Ceilometer Configuration Options

The meter pipeline definition looks like:

---
sources:
  - name: 'source name'
    meters:
      - 'meter filter'
    sinks:
      - 'sink name'
sinks:
  - name: 'sink name'
    transformers: 'definition of transformers'
    publishers:
      - 'list of publishers'

There are several ways to define the list of meters for a pipeline source. The list of valid meters can be found in Measurements. There is a possibility to define all the meters, or just included or excluded meters, with which a source should operate:

To include all meters, use the * wildcard symbol. It is highly advisable to select only the meters that you intend on using to avoid flooding the metering database with unused data.
To define the list of meters, use either of the following:
- To define the list of included meters, use the meter_name syntax.
- To define the list of excluded meters, use the !meter_name syntax.

Note

The OpenStack Telemetry service does not have any duplication check between pipelines, and if you add a meter to multiple pipelines then it is assumed the duplication is intentional and may be stored multiple times according to the specified sinks.

The above definition methods can be used in the following combinations:

Use only the wildcard symbol.
Use the list of included meters.
Use the list of excluded meters.
Use wildcard symbol with the list of excluded meters.

Note

At least one of the above variations should be included in the meters section. Included and excluded meters cannot co-exist in the same pipeline. Wildcard and included meters cannot co-exist in the same pipeline definition section.

The transformers section of a pipeline sink provides the possibility to add a list of transformer definitions. The available transformers are:

Name of transformer	Reference name for configuration
Accumulator	accumulator
Aggregator	aggregator
Arithmetic	arithmetic
Rate of change	rate_of_change
Unit conversion	unit_conversion
Delta	delta

The publishers section contains the list of publishers, where the samples data should be sent after the possible transformations.

Similarly, the event pipeline definition looks like:

---
sources:
  - name: 'source name'
    events:
      - 'event filter'
    sinks:
      - 'sink name'
sinks:
  - name: 'sink name'
    publishers:
      - 'list of publishers'

The event filter uses the same filtering logic as the meter pipeline.

Transformers¶

Note

Transformers maintain data in memory and therefore do not guarantee durability in certain scenarios. A more durable and efficient solution may be achieved post-storage using solutions like Gnocchi.

The definition of transformers can contain the following fields:

name: Name of the transformer.
parameters: Parameters of the transformer.

The parameters section can contain transformer specific fields, like source and target fields with different subfields in case of the rate of change, which depends on the implementation of the transformer.

The following are supported transformers:

Rate of change transformer¶

Transformer that computes the change in value between two data points in time. In the case of the transformer that creates the cpu_util meter, the definition looks like:

transformers:
    - name: "rate_of_change"
      parameters:
          target:
              name: "cpu_util"
              unit: "%"
              type: "gauge"
              scale: "100.0 / (10**9 * (resource_metadata.cpu_number or 1))"

The rate of change transformer generates the cpu_util meter from the sample values of the cpu counter, which represents cumulative CPU time in nanoseconds. The transformer definition above defines a scale factor (for nanoseconds and multiple CPUs), which is applied before the transformation derives a sequence of gauge samples with unit %, from sequential values of the cpu meter.

The definition for the disk I/O rate, which is also generated by the rate of change transformer:

transformers:
    - name: "rate_of_change"
      parameters:
          source:
              map_from:
                  name: "disk\\.(read|write)\\.(bytes|requests)"
                  unit: "(B|request)"
          target:
              map_to:
                  name: "disk.\\1.\\2.rate"
                  unit: "\\1/s"
              type: "gauge"

Unit conversion transformer¶

Transformer to apply a unit conversion. It takes the volume of the meter and multiplies it with the given scale expression. Also supports map_from and map_to like the rate of change transformer.

Sample configuration:

transformers:
    - name: "unit_conversion"
      parameters:
          target:
              name: "disk.kilobytes"
              unit: "KB"
              scale: "volume * 1.0 / 1024.0"

With map_from and map_to:

transformers:
    - name: "unit_conversion"
      parameters:
          source:
              map_from:
                  name: "disk\\.(read|write)\\.bytes"
          target:
              map_to:
                  name: "disk.\\1.kilobytes"
              scale: "volume * 1.0 / 1024.0"
              unit: "KB"

Aggregator transformer¶

A transformer that sums up the incoming samples until enough samples have come in or a timeout has been reached.

Timeout can be specified with the retention_time option. If you want to flush the aggregation, after a set number of samples have been aggregated, specify the size parameter.

The volume of the created sample is the sum of the volumes of samples that came into the transformer. Samples can be aggregated by the attributes project_id, user_id and resource_metadata. To aggregate by the chosen attributes, specify them in the configuration and set which value of the attribute to take for the new sample (first to take the first sample’s attribute, last to take the last sample’s attribute, and drop to discard the attribute).

To aggregate 60s worth of samples by resource_metadata and keep the resource_metadata of the latest received sample:

transformers:
    - name: "aggregator"
      parameters:
          retention_time: 60
          resource_metadata: last

To aggregate each 15 samples by user_id and resource_metadata and keep the user_id of the first received sample and drop the resource_metadata:

transformers:
    - name: "aggregator"
      parameters:
          size: 15
          user_id: first
          resource_metadata: drop

Accumulator transformer¶

This transformer simply caches the samples until enough samples have arrived and then flushes them all down the pipeline at once:

transformers:
    - name: "accumulator"
      parameters:
          size: 15

Multi meter arithmetic transformer¶

This transformer enables us to perform arithmetic calculations over one or more meters and/or their metadata, for example:

memory_util = 100 * memory.usage / memory

A new sample is created with the properties described in the target section of the transformer’s configuration. The sample’s volume is the result of the provided expression. The calculation is performed on samples from the same resource.

Note

The calculation is limited to meters with the same interval.

Example configuration:

transformers:
    - name: "arithmetic"
      parameters:
        target:
          name: "memory_util"
          unit: "%"
          type: "gauge"
          expr: "100 * $(memory.usage) / $(memory)"

To demonstrate the use of metadata, the following implementation of a novel meter shows average CPU time per core:

transformers:
    - name: "arithmetic"
      parameters:
        target:
          name: "avg_cpu_per_core"
          unit: "ns"
          type: "cumulative"
          expr: "$(cpu) / ($(cpu).resource_metadata.cpu_number or 1)"

Note

Expression evaluation gracefully handles NaNs and exceptions. In such a case it does not create a new sample but only logs a warning.

Delta transformer¶

This transformer calculates the change between two sample datapoints of a resource. It can be configured to capture only the positive growth deltas.

Example configuration:

transformers:
    - name: "delta"
      parameters:
        target:
            name: "cpu.delta"
        growth_only: True

Publishers¶

The Telemetry service provides several transport methods to transfer the data collected to an external system. The consumers of this data are widely different, like monitoring systems, for which data loss is acceptable and billing systems, which require reliable data transportation. Telemetry provides methods to fulfill the requirements of both kind of systems.

The publisher component makes it possible to save the data into persistent storage through the message bus or to send it to one or more external consumers. One chain can contain multiple publishers.

To solve this problem, the multi-publisher can be configured for each data point within the Telemetry service, allowing the same technical meter or event to be published multiple times to multiple destinations, each potentially using a different transport.

The following publisher types are supported:

gnocchi (default)¶

When the gnocchi publisher is enabled, measurement and resource information is pushed to gnocchi for time-series optimized storage. Gnocchi must be registered in the Identity service as Ceilometer discovers the exact path via the Identity service.

More details on how to enable and configure gnocchi can be found on its official documentation page.

panko¶

Event data in Ceilometer can be stored in panko which provides an HTTP REST interface to query system events in OpenStack. To push data to panko, set the publisher to panko://.

notifier¶

The notifier publisher can be specified in the form of notifier://?option1=value1&option2=value2. It emits data over AMQP using oslo.messaging. Any consumer can then subscribe to the published topic for additional processing.

The following customization options are available:

per_meter_topic

The value of this parameter is 1. It is used for publishing the samples on additional metering_topic.sample_name topic queue besides the default metering_topic queue.

policy

Used for configuring the behavior for the case, when the publisher fails to send the samples, where the possible predefined values are:

default: Used for waiting and blocking until the samples have been sent.
drop: Used for dropping the samples which are failed to be sent.
queue: Used for creating an in-memory queue and retrying to send the samples on the queue in the next samples publishing period (the queue length can be configured with max_queue_length, where 1024 is the default value).

topic

The topic name of the queue to publish to. Setting this will override the default topic defined by metering_topic and event_topic options. This option can be used to support multiple consumers.

udp¶

This publisher can be specified in the form of udp://<host>:<port>/. It emits metering data over UDP.

file¶

The file publisher can be specified in the form of file://path?option1=value1&option2=value2. This publisher records metering data into a file.

Note

If a file name and location is not specified, the file publisher does not log any meters, instead it logs a warning message in the configured log file for Telemetry.

The following options are available for the file publisher:

max_bytes: When this option is greater than zero, it will cause a rollover. When the specified size is about to be exceeded, the file is closed and a new file is silently opened for output. If its value is zero, rollover never occurs.
backup_count: If this value is non-zero, an extension will be appended to the filename of the old log, as ‘.1’, ‘.2’, and so forth until the specified value is reached. The file that is written and contains the newest data is always the one that is specified without any extensions.

http¶

The Telemetry service supports sending samples to an external HTTP target. The samples are sent without any modification. To set this option as the notification agents’ target, set http:// as a publisher endpoint in the pipeline definition files. The HTTP target should be set along with the publisher declaration. For example, additional configuration options can be passed in: http://localhost:80/?option1=value1&option2=value2

The following options are availble:

timeout: The number of seconds before HTTP request times out.
max_retries: The number of times to retry a request before failing.
batch: If false, the publisher will send each sample and event individually, whether or not the notification agent is configured to process in batches.
verify_ssl: If false, the ssl certificate verification is disabled.

The default publisher is gnocchi, without any additional options specified. A sample publishers section in the /etc/ceilometer/pipeline.yaml looks like the following:

publishers:
    - gnocchi://
    - panko://
    - udp://10.0.0.2:1234
    - notifier://?policy=drop&max_queue_length=512&topic=custom_target

Pipeline Partitioning¶

Note

Partitioning is only required if pipelines contain transformations. It has secondary benefit of supporting batching in certain publishers.

On large workloads, multiple notification agents can be deployed to handle the flood of incoming messages from monitored services. If transformations are enabled in the pipeline, the notification agents must be coordinated to ensure related messages are routed to the same agent. To enable coordination, set the workload_partitioning value in notification section.

To distribute messages across agents, pipeline_processing_queues option should be set. This value defines how many pipeline queues to create which will then be distributed to the active notification agents. It is recommended that the number of processing queues, at the very least, match the number of agents.

Increasing the number of processing queues will improve the distribution of messages across the agents. It will also help batching which minimises the requests to Gnocchi storage backend. It will also increase the load the on message queue as it uses the queue to shard data.

Warning

Decreasing the number of processing queues may result in lost data as any previously created queues may no longer be assigned to active agents. It is only recommended that you increase processing queues.

updated: 2020-06-08 01:26

Except where otherwise noted, this document is licensed under Creative Commons Attribution 3.0 License. See all OpenStack Legal Documents.