Fluent Commerce Docs

Get Started
Functionality
Learning
Extension
Dev Tooling
APIs
Releases

Content in this track:

Fluent Big Inventory - Features Overview

Getting Started with Fluent Big Inventory

How to Submit Inventory Batches

Metrics usage for Platform Observability

End Of Track

Content in this track: Metrics usage for Platform Observability

Essential knowledge

Authors:

Kirill Gaiduk, Yulia Andreyanova

Changed on:

19 June 2024

Overview

This document is intended to describe to implementers which Metrics are available within the Fluent Big Inventory product and what kind of data they provide to achieve Platform Observability.Pre-requisites:

You should have knowledge of How Metrics works.
You should have knowledge of Events.
You should have knowledge of Fluent APIs.

Key points

Metrics are available for all customers with Fluent Order Management and Fluent Big Inventory.
Metrics data is captured, uploaded, and stored in the Metrics workspace.
Metrics data is fetched and visualized for the Platform Observability enablement.

What is Metrics?

Metrics are numerical measurements that are:

Captured by instrumenting the source code.

We instrument the source code of the:

Fluent Platform Applications:
- Fluent APIs (accepting external HTTP API requests from the customer side);
- Orchestration Engine (Rubix) (executing orchestration event requests).
  
  Find the Metrics Capturing flow Explanation through an Example section below.
Batch Pre-Processing and Inventory Feeds jobs:
- Inventory Deduplication (accepting batch inventory upload requests);
- Inventory Data Loading (executing latest inventory data export to external storage).

Sent through a telemetry data manager to be stored in a monitoring system.
Saved with time series data format that refers to recording changes over time.

Info

The retention time for the ingested Metrics data is 150 days. After this period, the data is deleted from the account-specific Metrics workspace.

Visualized to provide visibility of the Platform operational data and allow to monitor and observe Platform / Inventory availability and orchestration insights.

Note

For more details, check the Fluent Big Inventory Web App Overview.

Info

What users want to measure differs from application to application. For an orchestration engine, it could be the number of requests received, time spent in the queue, and the actual processing time; for a data export job, it could be the number of loaded records, the completion timestamp, and so on.Metrics are important in understanding why your application works a certain way. Let's assume you are running a web application and discover it is slow. You will need some information to learn what is happening with your application. For example, when the number of requests is high, the application may become slow. If you have the request count Metric, you can determine the cause and improve the processing flow to handle the load.

Metrics vs. Events

An Event is typically used to log a specific activity of what has happened within the system.While the Metrics (or Metrics API) themselves do not capture events, they provide a capability to reflect an aggregated view for a range of captured events, which include:

Cross-retailer events,
Events initiated through GraphQL mutations (create events),
Cross workflow events,
User action events via the sync endpoint,
Events instigated through the async endpoint,
Scheduled events.

These Metrics are associated with the following endpoints:

POST /api/v4.1/event/async
POST /api/v4.1/event/sync
POST /api/v4.1/job/{jobId}/batch

Info

This scope does not include Metrics related to events categorized as "Action" (sent directly to the Event API).
Inline events triggered via workflow for a given orchestrated entity are not captured.

Types

The following Metric types are currently available:

Counter
A cumulative Metric representing a monotonically increasing counter. The value can only increase or be reset to zero upon process restart.
Gauge
Represents a single numerical value that can vary up and down.
Histogram
Samples observations (e.g., request durations, response sizes) and counts them in configurable buckets. Offers insights into data distribution.

Typical Usage Examples

Counter Metrics measures the following:
- total number of tasks completed,
- resource consumption,
- errors encountered.
Gauge Metrics measures fluctuating values, like:
- the timestamp of when the last event was received,
- or when the given operation has been completed.
Histogram Metrics expose multiple time series during a scrape:
- <histogram_metric>_bucket demonstrates a range, showing the observations less than or equal to its limit. For example, for the Metric indicating event runtime, you might have buckets like `<0.005`, `<0.01`, `<0.1` indicating how many events had run times less than those thresholds.
  The default bucket boundaries that are configured for histograms are measured in milliseconds and are as follows: `5ms`, `10ms`, `25ms`, `50ms`, `75ms`, `100ms`, `250ms`, `500ms`, `750ms`, `1000ms` (or `1s`), `2500ms` (or `2.5s`), `5000ms` (or `5s`), `7500ms` (or `7.5s`), `10000ms` (or `10s`) `+Inf` (which captures all values greater than the highest specified boundary).
- <histogram_metric>_count displays the total count of observed events and could be used when you're interested in the frequency of those events.
- <histogram_metric>_sum represents the total sum of all observed values and could be used when interested in the total value of events.
  
  When combined, the _sum and _count Metrics allow you to compute the average of the observed values. By dividing the _sum by _count, you can determine the mean of your observations. This provides a clearer understanding of the overall behaviour of the values being observed and the distribution given by the buckets.

Labels

Labels are saved as a part of the specific Metric to enable Metrics data filtering when fetched with the Metrics API queries. Label examples include (a list of all Metrics and supported labels can be found in the Available Metrics section below):

account_id
Fluent account identifier.
retailer_id
Fluent retailer identifier (within the account).
event_name
Describes the event name.
entity_type
Classifies the entity being referred to in the event.

Info

This includes Fluent domain entities, for example, `PRODUCT_CATALOGUE`, `INVENTORY_CATALOGUE`, `INVENTORY_POSITION`, `ORDER`, `FULFILMENT`, and `LOCATION`.

source
Specifies the origin of the event or the medium through which the event was channelled.

Source Determination Principals

A `"source"` value is set in accordance with the following patterns:1. External Events.If the event is externally generated, it may already have the `"source"` field set, which denotes its origin. This allows users to track the event's source directly in entity dashboards.Example payload for `"INVENTORY_DELTA"` event with the `"source"` set to POS: No alt provided

This way of setting source is relevant for any event across any entity.2. Endpoints.

Sync Endpoint ( /api/v4.1/event/sync )
If the event's `"source"` field is not set and originates from the sync endpoint, the source is set to event-sync.
Async Endpoint ( /api/v4.1/event/async )
For events coming from the async endpoint without a predefined `"source"`, it's labelled as event.
Batch Endpoint ( /api/v4.1/job/{jobId}/batch )
Events from the batch endpoint are designated with the batch `"source"`.

3. Scheduled and Cross Domain Events.
These events automatically receive the `"source"` value internal since they originate within the platform.

le (specific to Histograms)
`le` stands for "less than or equal to". It labels the buckets in a histogram and represents the upper bound of each bucket.

Info

For example, an `le` value of "5" in the context of a duration histogram would count all events that took 5 seconds or less.

Available Metrics

The following Metrics are currently available:

Fluent API Metrics

Name	Type	Labels	Description
core_event_received_total	Counter	account_id, retailer_id, event_name, entity_type, source	The number of events received by the Fluent platform.
core_event_last_received_seconds	Gauge	account_id, retailer_id, event_name, entity_type, source	The last timestamp of events received by the Fluent platform.

Orchestration Engine (Rubix Core/Batch) Metrics

Name	Type	Labels	Description
rubix_event_received_total	Counter	account_id, retailer_id, event_name, entity_type, source	The number of events received by the orchestration engine (Rubix) from the queue or direct HTTP API.
rubix_event_inflight_latency_seconds	Histogram	account_id, retailer_id, event_name, entity_type, source	The amount of time an event stays in the internal queue before Rubix receives it.
rubix_event_inflight_latency_seconds_sum	Counter	account_id, retailer_id, event_name, entity_type, source	Total accumulated time of events in the internal queue before Rubix receives them.
rubix_event_inflight_latency_seconds_count	Counter	account_id, retailer_id, event_name, entity_type, source	The number of observed events in the internal queue before Rubix receives them.
rubix_event_inflight_latency_seconds_bucket	Counter	account_id, retailer_id, event_name, entity_type, source, le	Bucketed counts of event latencies in the internal queue.
rubix_event_runtime_seconds	Histogram	account_id, retailer_id, event_name, entity_type, source, status	The amount of time Rubix takes to execute an event.
rubix_event_runtime_seconds_sum	Counter	account_id, retailer_id, event_name, entity_type, source, status	Total accumulated time taken by Rubix to execute events.
rubix_event_runtime_seconds_count	Counter	account_id, retailer_id, event_name, entity_type, source, status	The number of observed events Rubix executed.
rubix_event_runtime_seconds_bucket	Counter	account_id, retailer_id, event_name, entity_type, source, status, le	Bucketed counts of event runtimes by Rubix.

Batch Pre-Processing Metrics

The Metrics are captured from the Inventory Deduplication job:

Name	Type	Labels	Description
bpp_records_processed_total	Counter	account_id, run_id, stage, first_batch_received, deduplication_finished	The total amount of all batch items processed by the Inventory Deduplication job.The `first_batch_received` label is used to indicate the date of the first batch.The `deduplication_finished` label is used to indicate the end date of the Inventory Deduplication job execution.
bpp_records_unchanged_total	Counter	account_id, run_id, stage	The total amount of batch items that were filtered out by the Inventory Deduplication job.
bpp_records_changed_total	Counter	account_id, run_id, stage	The total amount of batch items that were considered “changed” and sent to Rubix by the Inventory Deduplication job.
bpp_last_run_timestamp_seconds	Gauge	account_id, run_id, stage, status	The timestamp for when the given Batch Pre-Processing operation has been completed.Status: `SUCCESS`, `ERROR`.

Inventory Feeds Metrics

The Metrics are captured from the Inventory Data Loading job:

Name	Type	Labels	Description
feed_sent_total	Counter	account_id, feed_ref, run_id, data_type	The total number of records exported by the Inventory Data Loading job, split by data type.Data type: `INVENTORY_POSITION`, `INVENTORY_CATALOGUE`, `VIRTUAL_POSITION`, `VIRTUAL_CATALOGUE`.
feed_last_run_timestamp_seconds	Gauge	account_id, feed_ref, run_id, status	The timestamp for when the Inventory Data Loading job completed.Status: `SUCCESS`, `ERROR`, `NO_RECORDS`. `NO_RECORDS` refers to the scenario where the job is triggered but no records have been updated, so there is nothing to export from the feed.

Metrics Capturing

The Metrics capturing is a foundation and a starting point for building a reliable data source to enable Platform Observability. The following example for Inventory Update end-to-end flow (via the customer’s request based on Point Of Sale (POS) data) is intended to serve as a reference for Metrics Capturing Process.

Explanation through an Example

The correspondent Metrics (the number of events received and their runtime) are logged upon:

Receiving of an Inventory Update customer’s request (based on the POS data) by Fluent Platform.
Starting the Inventory Catalogue Workflow for proceeding with the latest Inventory updates (e.g. reflecting a sale transaction) to ensure up-to-date On Hand quantities.
Starting the Virtual Catalogue Workflow for updating Available to Sell quantities.

The given example is one of the use cases.

The Metrics capturing is used in a wide range of the Fluent Platform Observability scenarios. For Example:

Batch Inventory Updates including the Batch Pre-Processing;
Inventory Updates upon Order reservations, processing or cancelling;
Inventory Feeds executions for Inventory Availability data exporting to external storage;
And many more.

Content in this track:

Fluent Big Inventory - Features Overview

Getting Started with Fluent Big Inventory

Inventory Ingestion

Inventory Batch

How to Submit Inventory Batches

Fluent Big Inventory Web App

Metrics Overview

How Metrics works

Metrics usage for Platform Observability

Inventory Feeds

How do Inventory Feeds work

End Of Track

Metrics usage for Platform Observability

Metrics usage for Platform Observability

Overview

Key points

What is Metrics?

Metrics vs. Events

Types

Labels

Available Metrics

Fluent API Metrics

Orchestration Engine (Rubix Core/Batch) Metrics

Batch Pre-Processing Metrics

Inventory Feeds Metrics

Metrics Capturing

Explanation through an Example

Related content

Metrics Overview

How Metrics works

Getting Started with Metrics API

Kirill Gaiduk

Contributors:

Knowledge Tracks Related To This Topic

Building Blocks

By Type

Helpful Resources

Quick Links