Azure Service Bus support

Azure Service Bus Support for Standard and Premium Namespaces

Last reviewed by Tyler Eastridge

AceMQ supports Azure Service Bus 24/7 across Americas, EMEA and APAC: Standard and Premium namespaces, queues, topics, sessions and the .NET, Java, Python and Node clients on AKS, App Service and Functions. Most calls start with one of three problems: lock expiry causing duplicates, Standard-tier throttling under bursts, or dead-letter queues growing without an application error.

AceMQ supports Azure Service Bus in production — lock expiry causing duplicate processing, Standard-tier throttling under burst load, dead-letter queues filling from MaxDeliveryCount, and Premium namespaces sized wrong for the actual message profile. Every ticket reaches a named senior engineer who already knows your namespace and entity layout.

Support quotes answered within 1 hour, by a senior engineer.

Planning a build, upgrade or migration instead? See AceMQ Azure Service Bus consulting and services.

160+
Customers supported
10+
Years in production support
40+
Sev 1 incidents handled a year
24/7
Americas, EMEA and APAC
1hour
Quote reply

Customers

  • Microsoft
  • Henry Schein
  • Enbridge
  • Gojek

Partners

  • Broadcom
  • Redis
  • VMware Tanzu
  • OSSeva

Incident Triage

Problems we fix every week on Azure Service Bus

Nine Service Bus failures we diagnose often, each with the metric, dead-letter reason or entity setting an engineer reads first.

Common Azure Service Bus production problems and how AceMQ resolves them
Symptom you're seeingWhat we check first
Messages are processed twice in PeekLock modeA received message is locked for the entity's LockDuration. If the handler is still running when the lock lapses, the broker makes the message available again and the original CompleteMessageAsync fails with MessageLockLostException. We compare handler p99 with the lock duration, check whether the client auto-renews the lock (MaxAutoLockRenewalDuration in the .NET processor, and the equivalent in the Java and Python SDKs), and look at prefetch. A large prefetch count means messages sit locked in client memory while earlier ones are still being handled, so the lock clock runs on messages nobody has started. The fix is a lock duration and renewal policy set from measured duration, a prefetch sized to the handler, and an idempotent handler.
ServerBusyException and throttled requests on StandardStandard tier throttles on a credit-based model, with a published operations-per-second figure of 1,000 for Basic and Standard (Microsoft Learn quotas page). Premium has no fixed operations-per-second limit; throughput follows the messaging units you allocate and the workload. We read the ThrottledRequests, IncomingRequests and ServerErrors metrics against the traffic shape. Typical causes are many small sends where batching would do, management operations in a hot loop (creating clients per message), and topic filter evaluation on a subscription-heavy topic. The remedy is batching, long-lived client reuse, trimming filters, or a measured move to Premium.
Premium namespace looks fine on traffic but throttlesPremium throttles on resource pressure. We read NamespaceCpuUsage and NamespaceMemoryUsage and the pending checkpoint count, then compare them with the number of messaging units and partitions (1, 2 or 4, fixed at creation). A namespace sized for average load can hit a CPU ceiling on a burst. Sizing is evidence-based: peak message rate, message size, number of subscriptions and filter complexity.
The dead-letter queue grows and the logs show nothingRead the dead-letter reason on the message, not only the count. MaxDeliveryCountExceeded (default delivery count 10) means the handler abandoned or crashed on it. TTLExpiredException means it aged out before a consumer got to it. A message that fails a subscription filter action or exceeds the header size ends there too. We trace each reason class back to a code path or a setting, set an alarm on dead-letter message count per entity, and write the replay procedure so nobody resubmits 40,000 messages into a live queue.
Queue depth climbs and consumers appear idleCheck four things. On a session-enabled queue, messages in a session can only go to the receiver that holds that session, so sessions with no receiver sit untouched while other receivers idle. Compare ActiveMessageCount with ScheduledMessageCount: scheduled messages are not yet deliverable and are easy to mistake for a stalled backlog. Look for a receiver stuck in a long wait behind a network proxy that drops idle AMQP links. And look for a poisoned session. Application health checks often report green throughout.
QuotaExceededException on sendThe entity has reached its maximum size. The configured size ranges from 1 to 5 GB, or up to 80 GB with partitioning on Basic and Standard, and 80 GB natively on Premium (Microsoft Learn quotas page). A large message over 1 MB sent to a queue is counted twice against size, and to a topic counted once for the topic plus once for each matching subscription. A stalled subscription fills the topic for everyone. We find the slow subscription, clean it up and set size alarms with room to act.
Subscribers miss messages that were publishedTopic filters. SQL filters (2,000 per topic) and correlation filters (100,000 per topic) both exist. A new subscription with no rule gets the default TrueFilter, but a rule replaced by a mistyped property name gets nothing. We dump the rules, test them against the real message properties, and check whether the publisher sets the property as a system or user property.
Connection drops from AKS or App ServiceAMQP connection limits per namespace are 5,000 concurrent connections, and 5,000 concurrent receive requests per entity (Microsoft Learn quotas page). Creating a new client per request, SNAT port exhaustion behind a NAT gateway, and idle-link timeouts through firewalls all show up as intermittent disconnects. We check connection counts in ActiveConnections, client lifetime, and the retry policy.
Geo-disaster recovery completed and consumers stayed downGeo-DR pairs two Premium namespaces behind an alias and fails the alias over. It carries entity metadata, not the messages sitting in the primary, so recovery means consumers reconnect through the alias and unprocessed work in the old primary is handled by plan. We rehearse the failover, including the point where applications still hold the old connection string.

Following any P1 we deliver a written cause that names the entity, setting or client behavior responsible.

Messages piling up in a dead-letter queue with no explanation? Request a quote and an engineer replies within the hour.

Contact Us

Deployment coverage

Which Azure Service Bus setups does AceMQ support, and who covers what?

SetupWhat AceMQ coversWhat Microsoft coversExcluded
Standard namespacesThrottling analysis, batching, lock and prefetch tuning, dead-letter handling, topic filter design, cost per operation reviewThe Azure Service Bus service and its SLA, quotas and platform faultsYour application code is reviewed, not rewritten
Premium namespacesMessaging-unit and partition sizing, CPU and memory pressure analysis, large messages over AMQP, private endpoints, Geo-DR rehearsalsThe service, including platform maintenance and capacity you have allocatedAzure regional or platform outages
Client SDKs (.NET, Java, Python, Node)Retry policy, connection reuse, auto-lock renewal, session handling, transaction useThe SDK as a Microsoft product when a defect is confirmedThird-party wrappers and frameworks not maintained upstream
AKS, Container Apps, App Service and Functions consumersScaling rules (KEDA, Functions trigger concurrency), graceful shutdown, connection countsThe Azure compute platformsCluster control-plane faults: see our Kubernetes support
JMS 2.0 over AMQP (Premium)JMS application behaviour on Service Bus, connection factory and session handlingThe AMQP endpointJMS providers other than the Service Bus library
Private endpoints and VNet-isolated namespacesDNS resolution, firewall and NSG rule review, identity and RBAC for managed identitiesAzure networking and Entra ID platformYour corporate network and proxy devices
Hybrid on-premises to AzureBridges, shovels, and migration paths between on-premises brokers and Service BusAzure sideOn-premises broker vendors' own products where unsupported by us

Support options

Who provides enterprise support for Azure Service Bus? Microsoft and independent support compared

Microsoft supports Azure Service Bus as part of its Azure support plans, and it owns the service, its SLA and its roadmap. It does not review the consumer code or entity design that cause most incidents. Independent support from AceMQ covers that layer, alongside whichever Azure plan you already hold.

Microsoft Azure support plans
AceMQ independent support
What it coversThe Azure service: availability, quotas and confirmed platform issues, across all Azure products
What it coversHow your producers, consumers and entities use Service Bus, plus tier and namespace design
PlansBasic (included), Developer ($29 a month), Standard ($100 a month), Professional Direct ($1,000 a month), Unified Enterprise (price not shown)
PlansOne contract; plan set by SLA, scope and deployment
HoursDeveloper: business hours. Standard, Professional Direct, Unified Enterprise: 24x7
Hours24/7 across Americas, EMEA and APAC
Fastest first responseSeverity A (critical) 1 hour on Standard and Professional Direct; Unified Enterprise lists 15 minutes for a critical system down on Azure
Fastest first responseResponse targets from 15 minutes, by plan
Root cause analysisPlatform-level
Root cause analysisWritten root cause analysis on every P1
PricePlan fee set by Microsoft
Price$24,000 to $72,000 a year (USD), set by SLA, scope and deployment. Quote reply within 1 hour
Beyond AzureAzure and Microsoft products
Beyond AzureThe same engineers also support RabbitMQ, Kafka, ActiveMQ and IBM MQ, which helps in a migration

Dated source: Azure support plans page, azure.microsoft.com/en-us/support/plans, read 17 November 2026. Short note under the table: raise a Microsoft case for a platform fault. Raise one with us when the symptom is in lock behaviour, throttling patterns, dead-letter growth or entity design.

What's Included

Everything in your Azure Service Bus support contract

Queues, topics and subscriptions, sessions, dead-letter handling, Geo-DR pairing and the client code that talks to them are covered in one agreement.

  • 01

    Emergency Incident Response

    Producers failing on quota, consumers stalled, or a namespace throttling in the middle of a business-critical window. With response targets from 15 minutes, by plan, a senior engineer joins a live bridge — not a ticket acknowledgement.

  • 02

    Root Cause Analysis

    Every P1 closes with a written RCA including the exception types involved, the metric evidence, the fix applied, and the entity or client change that prevents recurrence.

  • 03

    Tier & Throughput Sizing

    Standard versus Premium decisions backed by measured operation rates, messaging unit sizing, partitioning strategy, and prefetch and concurrency tuning against your real message sizes rather than a pricing page.

  • 04

    Client & SDK Debugging

    Azure.Messaging.ServiceBus, the legacy Microsoft.Azure.ServiceBus SDK, Java, Python, and JMS 2.0 clients — including the lock renewal, retry policy, and processor concurrency settings that cause most duplicate-delivery reports.

  • 05

    Identity, Networking & Compliance

    Managed identity and RBAC in place of SAS keys, private endpoints and service endpoint rules, customer-managed keys on Premium, and diagnostic settings that actually capture what you need during an incident.

  • 06

    Migration & Architecture Review

    Migration from on-premise MSMQ, IBM MQ, or ActiveMQ into Service Bus, plus honest guidance on when Event Hubs, Storage Queues, or a self-managed broker is the better fit for a given workload.

Why AceMQ

Why Azure teams bring in a messaging specialist

  • Named Engineers, Zero Cold Start

    The same senior engineers stay on your account. They know your namespaces, your entity layout, and which workloads sit behind which subscriptions — so a P1 call starts with diagnosis, not orientation.

  • No Tier-1 Triage Layer

    You reach a senior engineer directly by phone, email, or Slack. Nobody collects a repro to pass along, and there is no support-plan tier gate between you and someone who can read your Service Bus diagnostic logs.

  • The Client Side, Which Azure Support Won't Debug

    Most Service Bus incidents are client-side: lock renewal, prefetch, retry policy, or a handler that throws before Complete. Platform support scopes to the service, which means the layer causing your problem is the layer nobody is looking at.

  • Genuine Follow-the-Sun Coverage

    Round-the-clock coverage for customers in 26+ countries. Your 3am incident is someone's mid-afternoon — no overnight skeleton crew, no waiting for a region to wake up.

  • Proactive, Not Just Reactive

    Quarterly reviews of dead-letter growth, active message age, throttled-request trend, and entity size headroom — so you catch a subscription filter that stopped matching before it becomes a month of missing data.

  • We Know the Alternatives Too

    We support RabbitMQ, Kafka, IBM MQ, and ActiveMQ at the same depth. That means we will tell you when Service Bus is the wrong tool for a workload, and when Event Hubs or a self-managed broker would cost less and behave better.

Azure messaging setups we cover

Premium and Standard namespaces, called from AKS, App Service, Functions, Container Apps or VMs, including private endpoints, multi-region pairs and sovereign clouds.

  • Azure Service Bus Premium
  • Azure Service Bus Standard
  • Azure Kubernetes Service (AKS)
  • Azure App Service & Functions
  • Azure Container Apps
  • Azure VMs & VM Scale Sets
  • Hybrid on-premise to Azure
  • Multi-region with Geo-DR pairing
  • Private endpoints & VNet-isolated
  • Azure Government & sovereign clouds
  • .NET, Java, Python & Node clients
  • JMS 2.0 over AMQP

Version coverage

Versions we support (tiers, features and clients)

Azure Service Bus is managed, so there is no broker version to patch. What changes between estates is the tier, the features switched on and the client library.

Versions we support (tiers, features and clients)
AreaWhat we support
TiersBasic, Standard and Premium
Premium featuresMessaging units, namespace partitions (1, 2 or 4), large messages up to 100 MB over AMQP (default 1 MB), JMS 2.0, Geo-DR, private endpoints
Messaging featuresQueues, topics and subscriptions, sessions, scheduled messages, auto-forwarding, duplicate detection, dead-letter queues, transactions
Clients.NET (Azure.Messaging.ServiceBus), Java, Python and JavaScript Azure SDKs; JMS over AMQP
ProtocolsAMQP 1.0, and HTTPS where you need it (messages over HTTP are limited to 1 MB on Premium)

Quota facts (Microsoft Learn, "Azure Service Bus quotas and limits", page dated 13 October 2026, read 17 November 2026): Basic and Standard maximum message size 256 KB; Premium up to 100 MB over AMQP; 2,000 subscriptions per topic; 5,000 concurrent AMQP connections per namespace.

Escalation Path

Your first hour of an Azure Service Bus outage

The engineer on the line reads dead-letter reasons and namespace metrics without a briefing, so the call opens with the entity and the error code.

  1. From 15 min (by plan)first responseEmergencyProduction outage: a namespace unreachable, consumers down after a Geo-DR failover, every send failing with a quota or server-busy error
  2. 1 hourfirst responseCriticalSevere degradation: Premium throttling under CPU or memory pressure, queues growing while consumers idle, connections dropping from AKS or App Service
  3. 4 hoursfirst responseHighMessages processed twice after lock expiry, dead-letter queues growing with no application errors, subscribers missing messages after a filter change
  4. Next Dayfirst responseStandardQuestions on Standard versus Premium, session design, delivery count settings, JMS over AMQP and migration paths

Response targets scale with plan: a namespace throttling production traffic gets a faster clock than a pricing question.

Want named engineers who already know your namespaces, topics and filters? Ask for a quote or send a message.

Contact Us

Pricing

What does AceMQ Azure Service Bus support cost?

Azure Service Bus support typically costs $24,000–72,000 a year. The number of namespaces, whether you run Standard or Premium, how many messaging units and entities you manage, and the response plan set the price.

Quoted in US dollars (USD). Response targets start at 15 minutes by plan, 24/7, and quote requests are answered within 1 hour. Adding other brokers or technologies to the contract brings a multi-technology discount.

Support In Practice

Service Bus incidents we have resolved

Azure Service Bus support and assessment work from AceMQ. Each card links to the full write-up: the namespace and workloads, what AceMQ diagnosed or changed, and the result, so you can see how a similar problem was handled.

  • Architecture Advisory

    Azure Service Bus to RabbitMQ Migration for Financial Services

    Financial Services Firm

    A financial services firm is planning a migration from Azure Service Bus to a 3-node RabbitMQ cluster for improved compliance control and disaster recovery, sized for 1,300 msg/sec with 200KB payloads and warm schema replication DR.

    Azure Service BusRabbitMQ
  • Consulting

    Microsoft Service Bus to RabbitMQ Migration for Banking

    Banking Organization

    A bank dissatisfied with Microsoft Azure Service Bus performance and flexibility is evaluating a migration to RabbitMQ, with AceMQ proposing an assessment that compares total cost, operational control, and messaging capability for banking workloads.

    Azure Service BusRabbitMQ
All use cases

Support options

Four tiers of Azure Service Bus support.

AceMQ offers Azure Service Bus support in four tiers, each handing more of the day-to-day work to AceMQ: 24/7 support and incident management for namespaces your team runs, proactive monitoring that catches throttling, lock and dead-letter problems first, managed Service Bus inside your own Azure tenant, and fully managed Service Bus run on your behalf from AceMQ-managed Azure subscriptions and infrastructure, including the pieces around it. Standard and Premium namespaces are covered in all four.

  • Tier 1

    Azure Service Bus 24/7 Support & Incident Management

    Your developers own the apps, we take the hard faults

    Best for: Teams building on Service Bus who want a messaging engineer for lock, throttling and dead-letter problems that resist debugging.

  • Tier 2

    Azure Service Bus 24/7 Support with Proactive Monitoring

    Reviews of entities, locks, filters and quotas

    Best for: Teams that discover prefetch, lock duration and delivery count problems through duplicate orders instead of a review.

  • Tier 3

    Managed Azure Service Bus on Your Infrastructure

    We operate your messaging layer in your tenant

    Best for: Organizations whose namespaces must stay in their own Azure subscription, with specialists managing entities and monitoring.

  • Tier 4

    Fully Managed Azure Service Bus on AceMQ Infrastructure

    AceMQ runs your messaging operations as a service

    Best for: Teams that want reliable queues and topics without building an in-house Service Bus operations skill set.

FAQ

Azure Service Bus support questions

Still have questions about Azure Service Bus? Email an Expert

Who provides enterprise support for Azure Service Bus?

Microsoft supports the Azure Service Bus service under its Azure support plans. Independent providers such as AceMQ support how you use it: entity design, lock and prefetch settings, throttling, dead-letter handling, tier choice and client code behaviour. AceMQ works 24/7 with response targets from 15 minutes, by plan, and a senior engineer assigned to your account on every ticket. Many estates hold both.

What does 24/7 Azure Service Bus support cover with AceMQ?

It covers round-the-clock incident response, a written root cause on every P1, and standing work such as lock and delivery count reviews, quota checks and Geo-DR rehearsals. A Sev 1 like a namespace throttling every send starts within 15 minutes on the top plans.

Why are our Service Bus messages processed twice?

In PeekLock mode a message is locked for the lock duration. If the handler runs longer and the lock is not renewed, the message is delivered again and completing the first copy fails with MessageLockLostException. Large prefetch counts make it worse because locks start before handling does. We set lock duration and auto-renewal from measured handler time and make handlers idempotent.

How do we fix ServerBusyException on Azure Service Bus?

First find which tier you are on. Basic and Standard throttle on a credit model with a published 1,000 operations per second. Premium throttles on CPU and memory of the allocated messaging units. We read the throttled-request metrics against traffic, then batch sends, reuse clients and trim filters, or resize Premium. Moving tier is a measured decision, not a reflex.

Should we use Standard or Premium?

Premium gives isolated resources, predictable throughput, messages up to 100 MB over AMQP, JMS 2.0, Geo-DR and private-link features. Standard is shared and throttles. Choose Premium for production workloads with burst, size or isolation needs. We size it from your peak rate, message size, subscription count and filter complexity, so you pay for the units you need.

Why is our dead-letter queue filling with no application errors?

Read the dead-letter reason. MaxDeliveryCountExceeded means the message was delivered the maximum number of times (10 by default) without completing. TTLExpiredException means the message expired before a consumer received it. Other reasons are filter action errors and header size. Each points to a different fix, and we write a replay procedure so recovery does not flood the live entity.

What happens to messages during an Azure Service Bus Geo-DR failover?

Geo-disaster recovery pairs Premium namespaces behind an alias and fails the alias over. It replicates entity metadata, not the messages waiting in the primary, so you plan for unprocessed work and for clients that still hold the old connection string. We rehearse the failover before you need it.

Can AceMQ help us migrate onto or off Azure Service Bus?

Yes. We support both directions, using patterns such as bridged queues and dual consumption so cutover is reversible. We have worked on Service Bus to RabbitMQ migrations for financial services and banking clients, and on Service Bus cost reviews. See the linked use cases.

Does Azure Service Bus support JMS?

Yes, on Premium, which offers JMS 2.0 over AMQP. Java and Spring applications can then use Service Bus through standard JMS interfaces, and we help with connection factory settings, transactions and redelivery behavior.

Where does AceMQ support end and Microsoft support begin?

Microsoft runs the platform: availability, capacity, quotas and confirmed service defects. AceMQ works on your side of the line: entity design, client settings, lock handling, filters and diagnosing why a message behaved as it did.

Is a multi-year or annual commitment required for Service Bus help?

No. A throttling investigation, a dead-letter and lock review, a Geo-DR failover rehearsal or a migration plan can each be scoped as a separate job. If messaging incidents continue, an annual agreement adds 24/7 response and engineers who know your namespaces.

What affects the cost of Azure Service Bus support?

Annual plans run $24,000 to $72,000 (USD). We look at namespace count, Standard or Premium tier, messaging units, how many queues and topics carry production traffic, and the response level. Microsoft's service charges are separate. Describe the setup and a quote returns within 1 hour.

Let's talk

Managed Messaging Still Fails in Ways Only You Can Fix

Send your namespace tiers and consumer platforms. We reply to quote requests within 1 hour.

A senior Azure Service Bus engineer, not a sales queueQuote requests answered within 1 hourAd hoc engagements or an annual partnership

Production incident right now?Email an Azure Service Bus Architect