Rediger

Del via

Reliability in Azure Stream Analytics

Azure Stream Analytics is a fully managed real-time analytics service designed to process and analyze streaming data from multiple sources simultaneously, and lets you build complex event processing pipelines with SQL-like queries. Stream Analytics is a highly resilient service, with extensive redundancy and resiliency built in.

When you use Azure, reliability is a shared responsibility. Microsoft provides a range of capabilities to support resiliency and recovery. You're responsible for understanding how those capabilities work within all of the services you use, and selecting the capabilities you need to meet your business objectives and uptime goals.

This article describes Stream Analytics is resilient to a variety of potential outages and problems, including transient faults and availability zone outages. It also provides guidance about how to protect mission-critical jobs against region outages and service maintenance, and highlights some key information about the Stream Analytics service level agreement (SLA).

Important

When you consider the reliability of Stream Analytics, you also need to consider the reliability of your data sources, including inputs and outputs. Improving the resiliency of Stream Analytics alone might have limited impact if the other components aren't equally resilient. Depending on your resiliency requirements, you might need to make configuration changes across multiple areas.

Production deployment recommendations

To ensure high reliability in production environments with Stream Analytics, we recommend that you:

  • Use regions with availability zones: Deploy your streaming jobs and other resources in regions that support availability zones.
  • Deploy sufficient capacity: Set your streaming units based on your expected throughput, with additional capacity for handling peak loads, and with a buffer above your baseline requirements in case of sudden increases.
  • Monitor health: Implement comprehensive monitoring using Azure Monitor metrics and diagnostic logs to track job health, input/output events, and resource utilization. Configure alerts for critical metrics like watermark delay and runtime errors to detect issues before they impact data processing. For more information, see Monitor Azure Stream Analytics.
  • For mission-critical streaming workloads: Consider implementing a multi-region deployment strategy with synchronized job configurations across regions. While Stream Analytics doesn't provide native multi-region replication, you can achieve regional redundancy by deploying identical jobs in multiple regions with appropriate data routing mechanisms. For more information, see Custom multi-region solutions for resiliency.

Reliability architecture overview

This section describes some of the important aspects of how the service works that are most relevant from a reliability perspective. The section introduces the logical architecture, which includes some of the resources and features that you deploy and use. It also discusses the physical architecture, which provides details on how the service works under the covers.

Logical architecture

A streaming job, or simply job, is the fundamental unit in Stream Analytics that allows you to define and run your stream processing logic. A job consists of the following major components:

  • Inputs that read streaming data from data sources, such as Azure Event Hubs, Azure IoT Hub, or Azure Storage.
  • A query that processes and transforms the data.
  • Outputs that continuously write results to various destinations, such as Azure SQL Database, Azure Data Lake Storage, Azure Cosmos DB, Power BI, and more.

For more information on Stream Analytics jobs and the resource model, see Azure Stream Analytics resource model.

Physical architecture

Stream Analytics is designed for high reliability, with multiple layers of resiliency to any problems in the underlying infrastructure or input and output data sources. The following components help Stream Analytics to achieve robustness for your jobs:

  • Worker nodes. Stream Analytics jobs on virtual machines (VMs) called worker nodes that run within a cluster. When you use the Standard or StandardV2 SKUs, your jobs run on shared clusters. When you use the Dedicated SKU your jobs run on their own dedicated cluster.

    Because the platform automatically manages worker node creation, job placement across worker nodes, health monitoring, and the replacement of unhealthy worker nodes, you don't see or manage the VMs directly.

  • Streaming units. While the platform manages worker nodes and job distribution across worker nodes, you're responsible for allocating streaming units (SUs) to jobs. SUs represent the compute resources that are allocated to execute a job. The higher the number of SUs, the more compute resources are allocated for the job. For more information, see Understand and adjust Stream Analytics streaming units.

  • Checkpoints. Stream Analytics maintains job state through regular checkpointing of state. Checkpoints enable quick recovery with minimal data reprocessing in case of failures, even for jobs that use stateful query logic.

    When processing failures occur, Stream Analytics automatically restarts from the last checkpoint and reprocesses events that weren't fully processed. This guarantee applies to all built-in functions and user-defined functions within the job. However, achieving end-to-end exactly-once delivery depends on your output destination's capabilities. For more information, see Checkpoint and replay concepts in Azure Stream Analytics jobs.

Note

With Azure Stream Analytics on IoT Edge you can run jobs on your own infrastructure. When you use Stream Analytics on IoT Edge, you're responsible for configuring it to meet your reliability requirements. Stream Analytics on IoT Edge is outside the scope of this article.

Resilence to transient faults

Transient faults are short, intermittent failures in components. They occur frequently in a distributed environment like the cloud, and they're a normal part of operations. Transient faults correct themselves after a short period of time. It's important that your applications can handle transient faults, usually by retrying affected requests.

All cloud-hosted applications should follow the Azure transient fault handling guidance when they communicate with any cloud-hosted APIs, databases, and other components. For more information, see Recommendations for handling transient faults.

Stream Analytics automatically handles many transient faults for both ingesting data from inputs and writing data to outputs through built-in retry mechanisms. When a worker node running your job restarts, or if the job is moved between worker nodes, the job uses checkpoints to automatically replay any processing work it needs to do to catch up.

It's a good practice to configure output error policies. However, these policies only apply to data conversion errors, and they don't influence the behavior for handling transient faults.

Resilience to availability zone failures

Availability zones are physically separate groups of datacenters within an Azure region. When one zone fails, services can fail over to one of the remaining zones.

Stream Analytics is automatically zone-redundant in regions that support availability zones, which means jobs use multiple availability zones. Zone redundancy ensures that your job is resilient to a large set of failures, including catastrophic datacenter outages, without any changes to the application logic.

When you create a Stream Analytics job in a zone-enabled region, the service distributes your job's compute resources across multiple availability zones, as illustrated in the following diagram:

Diagram that shows a zone-redundant Stream Analytics job.

This zone-redundant deployment model ensures that your streaming jobs continue to process data even if an entire availability zone becomes unavailable. For example, the following diagram shows how jobs continue to run if zone 3 experiences an outage:

Diagram that shows a zone-redundant Stream Analytics job continuing to run when a zone is down.

Zone redundancy applies to all Stream Analytics features including query processing, checkpointing, and job management operations. Your job's state and checkpoint data are automatically replicated across zones, ensuring no data loss and near-zero downtime during zone failures.

Requirements

  • Region support: Zone redundancy for Stream Analytics resources is supported in any region that supports availability zones. For the complete list of regions that support availability zones, see Azure regions with availability zones.
  • SKUs: Zone redundancy is available in all Stream Analytics SKUs.

Cost

Zone redundancy on Stream Analytics doesn't incur additional charges. You pay the same rate for streaming units whether your job runs in a zone-redundant configuration or not. For information, see Azure Stream Analytics pricing.

Configure availability zone support

Behavior when all zones are healthy

This section describes what to expect when Stream Analytics jobs are configured with availability zone support and all availability zones are operational.

  • Traffic routing between zones. Stream Analytics runs each job on worker nodes. Incoming streaming data might be processed by workers in any zone. The service uses internal load balancing to distribute processing tasks across zones.

  • Data replication between zones. Stream Analytics replicates job state and checkpoint data synchronously across availability zones. When your job processes events and updates its state, these changes are written to multiple zones before being acknowledged. This synchronous replication ensures zero data loss even if an entire zone becomes unavailable. The replication process is transparent to your application and doesn't impact processing latency under normal conditions.

Behavior during a zone failure

This section describes what to expect when Stream Analytics jobs are configured with availability zone support and there's an availability zone outage.

  • Detection and response: The Stream Analytics platform is responsible for detecting a failure in an availability zone and responding. Workers in the failed zone are marked as unhealthy, and jobs that are running on those workers are automatically redistributed to workers in the remaining healthy zones. You don't need to do anything to initiate a zone failover.
  • Notification: Microsoft doesn't automatically notify you when a zone is down. However, you can use Azure Resource Health to monitor for the health of an individual resource, and you can set up Resource Health alerts to notify you of problems. You can also use Azure Service Health to understand the overall health of the service, including any zone failures, and you can set up Service Health alerts to notify you of problems.

  • Active requests: Running jobs are shifted to another worker in a healthy availability zone.

    Stream Analytics uses checkpointing to maintain processing state. During a zone failure, in-flight events being processed by workers in the failed zone are automatically reprocessed from the last checkpoint by workers in healthy zones.

  • Expected data loss: The job checkpointing system ensures no data loss.

  • Expected downtime: Jobs in progress automatically resume after the platform moves them to a healthy worker.

  • Traffic rerouting: The service automatically redirects all new input data to workers in healthy zones. Existing connections from input sources are re-established with workers in operational zones. Output connections are similarly re-established, ensuring continuous data flow through your streaming pipeline.

Zone recovery

When the failed availability zone recovers, Stream Analytics automatically reintegrates it into the active processing pool. Jobs begin to use the recovered infrastructure.

You don't take any action for zone recovery, because the platform handles all aspects of zone recovery operations including state synchronization and workload redistribution.

Test for zone failures

The Stream Analytics platform manages traffic routing, failover, and zone recovery. This feature is fully managed, so you don't need to initiate or validate availability zone failure processes.

Resilience to region-wide failures

Stream Analytics resources are deployed into a single Azure region. If the region becomes unavailable, your jobs (and dedicated clusters, if applicable) are also unavailable.

Custom multi-region solutions for resiliency

To achieve multi-region resilience for your streaming workloads, consider deploying separate jobs in multiple regions. When you do so, you're responsible for deploying and managing the jobs, and for configuring the appropriate data routing and synchronization strategies. The Stream Analytics jobs are two separate entities. It's the responsibility of your application to both send input data into the two regional inputs and reconcile between the two regional outputs. For more information about this approach, see Achieve geo-redundancy for Stream Analytics jobs.

Backup and restore

Stream Analytics doesn't have a built-in backup and restore feature.

However if you want to move, copy or back up the definition and configuration of your jobs, you can use the Stream Analytics extension for Visual Studio Code to export an existing job in the Azure cloud to your local computer. Once you save the entire configuration of your Stream Analytics jobs locally, you can then deploy it to the same or another Azure region. To learn how to copy, back up, and move your Stream Analytics jobs, see Copy, back up and move your Azure Stream Analytics jobs.

Resilience to service maintenance

Stream Analytics performs automatic platform maintenance to apply security updates, deploy new features, and improve service reliability. As a result, Stream Analytics can have service updates deployed on a weekly (or more frequent) basis. The Stream Analytics service ensures any new update passes rigorous internal rings to have the highest quality.

Consider the following points to ensure your jobs are resilient to service maintenance activities:

  • Configure jobs to be resilient to replays: Checkpoints are usually used to restore data after service maintenance. However, occasionally a replay technique needs to be used instead of a checkpoint. For more information and to learn how to configure your input data sources so that replays don't cause incorrect or partial results in your output, see Job recovery from a service upgrade.

  • Consider mitigating the risk of bugs by deploying identical jobs: The service proactively looks for many signals after deploying to each batch to get more confidence that there are no bugs introduced. However, no matter how much testing is done, there's still a risk that an existing, running job may break due to the introduction of a problem introduced by maintenance. If you are running mission-critical jobs, these risks need to be avoided.

    You can reduce the risk of a bug affecting your workload by deploying identical jobs to two Azure regions. You should then monitor these jobs to get notified when something unexpected happens. If one of these jobs ends up in a Failed state after a Stream Analytics service update, you should:

    • Contact Azure support to help identify the cause and resolve the problem.
    • Fail over any downstream consumers to use the healthy job output.

    When you select Azure regions to use for your secondary job, consider whether your region has a paired region. The Azure regions list has the most up-to-date information on which regions are paired. Stream Analytics guarantees that infrastructure in paired regions are updated at different times. The deployment of an update to Stream Analytics doesn't occur at the same time in a set of paired regions. As a result there is a sufficient time gap between the updates to identify potential issues and remediate them.

Service-level agreement

The service-level agreement (SLA) for Azure services describes the expected availability of each service and the conditions that your solution must meet to achieve that availability expectation. For more information, see SLAs for online services.

Stream Analytics provides separate availability SLAs for API calls to manage jobs, and for the operations of the jobs.

Related content

  • Last updated on