Amazon SQS support

Amazon SQS Support for Standard, FIFO and Lambda-Driven Queues

Last reviewed by Tyler Eastridge

AceMQ supports Amazon SQS 24/7 across Americas, EMEA and APAC: standard queues, FIFO queues, high-throughput FIFO and the Lambda, ECS and EKS consumers behind them. Most calls start with one of three problems: messages processed twice because the visibility timeout is shorter than the handler, a FIFO queue stuck behind one message group, or a dead-letter queue filling unnoticed.

AceMQ supports Amazon SQS in production — duplicate processing from visibility timeouts shorter than the handler, FIFO queues capped by a single message group, dead-letter queues that fill silently, and Lambda event source mappings that scale the wrong way under load. Every ticket reaches a named senior engineer who already knows your queue and consumer topology.

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

Planning a build, upgrade or migration instead? See AceMQ Amazon SQS 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 Amazon SQS

Duplicate deliveries, throttled FIFO queues, silent dead-letter queues and Lambda scaling puzzles, with the metric or setting we look at first.

Common Amazon SQS production problems and how AceMQ resolves them
Symptom you're seeingWhat we check first
The same message is processed by two workersCompare the handler's p99 duration with the queue's VisibilityTimeout (default 30 seconds, maximum 12 hours, per the AWS quotas page). If a handler can outlive the timeout, the message becomes visible again while the first worker is still busy. The fix is a timeout set from measured duration, ChangeMessageVisibility heartbeats for long jobs, and an idempotent consumer, because standard queues are at-least-once and a duplicate is allowed behaviour. For Lambda consumers, the AWS guidance is a queue timeout of at least six times the function timeout, plus the batch window. We check that number before anything else.
A FIFO queue is stuck at a few hundred messages a secondEach FIFO partition is limited to 300 transactions per second per API action in default mode, and up to 3,000 messages a second per action with 10-message batches (AWS quotas page). One constant MessageGroupId forces one lane. We read the group-ID distribution from the producer code, move the group key from "the whole system" to the entity that actually needs ordering (account, order, device), switch on high-throughput mode and batch in tens. Then we look at the consumer: with ordered groups, one failing message blocks everything behind it in the same group until it is deleted or dead-lettered.
The dead-letter queue is filling and nobody was pagedAlarm on ApproximateNumberOfMessagesVisible for every DLQ, not only on the source queue. The cause is almost always the redrive policy: maxReceiveCount missing, or set so high that one poison message cycles for hours. We also check the DLQ's retention. A message keeps its original enqueue timestamp when it moves, so a DLQ with the same 4-day retention as the source can delete a message that has only just arrived. Set the DLQ retention longer than the source. When you redrive from the DLQ, redrive at a controlled rate; an unthrottled redrive into a live queue is a classic P1.
The backlog grows while consumers look healthyWatch ApproximateAgeOfOldestMessage, not queue depth. Age tells you whether work is waiting too long. Then split in-flight messages (ApproximateNumberOfMessagesNotVisible) from waiting ones (ApproximateNumberOfMessagesVisible). High in-flight with a flat delete rate means workers are receiving and not finishing: look for a slow dependency, a timeout loop or a crashed pod holding messages until the visibility timeout expires.
Lambda is not scaling with the queue, or scales too hardCheck the event source mapping: batch size, MaximumBatchingWindowInSeconds, MaximumConcurrency, and whether the function returns partial batch failures with ReportBatchItemFailures. Without it, one bad record sends the whole batch back and the redelivery multiplies load. Look at function errors and throttles together with the queue metrics. A function that throttles on account concurrency makes the queue look like the problem.
Messages vanish between SNS or EventBridge and the queueThis is a policy problem, not a loss. Read the queue access policy for the publishing service principal and the aws:SourceArn condition. On an SSE-KMS queue, the key policy must let the service use the key. Check the SNS delivery-status logs and the target's DLQ. Nothing appears in your consumer logs, which is why teams spend days on the wrong side.
Throttling, empty receives and a bill that does not match the trafficSQS bills by request, so the same settings that cause incidents cause cost. Short polling returns empty responses even when messages exist; long polling with WaitTimeSeconds of 20 cuts empty receives and cost. Batch sends and deletes in tens. For standard queues that share one queue between many tenants, AWS now offers fair queues keyed on MessageGroupId (AWS quotas page); we assess whether that fits before you split queues by customer.
Large payloads and the extended clientThe maximum message size is 1 MiB, and the S3-backed Extended Client Library (Java and Python, synchronous clients only, up to 2 GB per payload) needs the consumer to resolve the pointer too. A producer using it against a plain consumer delivers an S3 reference that nobody reads.

A P1 closes with a written root cause naming the queue setting, policy or consumer behavior responsible and the fix applied.

Messages processed twice, or a dead-letter queue nobody was paged about? Request a quote and an engineer replies within the hour.

Contact Us

Deployment coverage

Which Amazon SQS setups does AceMQ support, and who covers what?

AceMQ supports the layer around the queue: how you send, receive, retry and secure. AWS runs the service itself. One contract can span several of the rows below.

SetupWhat AceMQ coversWhat AWS coversExcluded
SQS standard queuesVisibility timeout design, redrive policies, long polling, batching, idempotency review, CloudWatch alarms, cost per message reviewThe SQS service: availability, durability, quota increases and confirmed platform faultsYour application code is reviewed, not rewritten
SQS FIFO and high-throughput FIFOMessage group design, deduplication settings, throughput tuning, poison-message handling in ordered groupsService limits and the regional throughput quotasOrdering guarantees you did not design into the group key
Lambda event source mappingsBatch size, batching window, MaximumConcurrency, partial batch responses, queue timeout relative to function timeoutLambda platform behaviour and its quotasFunction code and runtime faults outside the agreed scope
ECS, Fargate and EKS pollersConsumer concurrency, graceful shutdown on SIGTERM so in-flight messages are released, autoscaling on queue ageThe container platform and its control planeCluster faults: see our Kubernetes support
SNS, EventBridge and S3 event sourcesQueue and key policies, subscription filter policies, delivery-status loggingDelivery behaviour of SNS and EventBridgeThird-party SaaS webhooks outside AWS
Encrypted and cross-account queuesSSE-KMS key policies, cross-account queue policies, VPC endpoint policiesKey management service availabilityKey custody decisions stay with your security team
AWS GovCloud and regulated accountsSame scope, inside your access and change controlsThe GovCloud serviceYour compliance attestations

Support options

Who provides enterprise support for Amazon SQS? AWS Support and independent support compared

AWS supports the SQS service itself under an AWS Support plan. It does not debug your consumer code, and SQS support is not sold as a standalone product. Independent support from AceMQ covers the consumer and integration layer where nearly every SQS incident sits, and sits alongside whichever AWS plan you hold.

AWS Support (Business Support+, Enterprise Support, Unified Operations)
AceMQ independent support
What it coversAWS services, including SQS, as a service: availability, quotas, confirmed platform issues
What it coversThe way your producers and consumers use SQS, on any AWS account structure
Hours24/7 phone, web, chat and email on the three plans listed on the AWS plans page
Hours24/7 across Americas, EMEA and APAC
Fastest first responseBusiness-critical system down: under 30 minutes (Business Support+), under 15 minutes (Enterprise Support); production system down under 1 hour on these plans
Fastest first responseResponse targets from 15 minutes, by plan
Scope of helpService behaviour and documented use; AWS lists consultative guidance on higher plans
Scope of helpConsumer code paths, retry and backoff, Lambda concurrency, IAM and KMS policies, cost per message
Root cause analysisPlatform-level
Root cause analysisWritten root cause analysis on every P1, with the CloudWatch evidence
PriceSet by the AWS plan
Price$24,000 to $72,000 a year (USD), set by SLA, scope and deployment. Quote reply within 1 hour
Multi-broker viewAWS services only
Multi-broker viewOne contract can also cover RabbitMQ, Kafka, ActiveMQ and IBM MQ

Dated source: AWS Support plans page, aws.amazon.com/premiumsupport/plans/, read 17 November 2026. Short note under the table: keep your AWS plan. Raise a case with AWS when the service itself misbehaves, and with us when the symptom sits in how the queue is used. Most estates use both.

What's Included

Everything in your Amazon SQS support contract

Producers, consumers, redrive policies, Lambda event source mappings, IAM and KMS policies around your queues are in scope, in as many accounts as you use.

  • 01

    Emergency Incident Response

    Queue backlog growing without bound, consumers stalled, or a redrive that pushed thousands of messages back into a live queue. 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: the CloudWatch evidence, the queue and consumer settings involved, the fix applied, and the change that prevents recurrence.

  • 03

    Throughput & Cost Tuning

    Long polling, batch sizing, visibility timeout calibration, FIFO message group design, and consumer concurrency tuned against actual handler duration. SQS bills per request, so tuning throughput and tuning cost are the same exercise.

  • 04

    Consumer & Integration Debugging

    Lambda event source mappings, ECS and Fargate pollers, Spring Cloud AWS, and the AWS SDKs across Java, Python, .NET, and Node — including the retry and backoff behavior that turns a transient fault into a DLQ flood.

  • 05

    IAM, KMS & Cross-Account Access

    Queue policy and IAM design, SSE-KMS key policies for SNS, EventBridge, and S3 event sources, VPC endpoint configuration, and the cross-account patterns where a silent access denial looks exactly like message loss.

  • 06

    Architecture Review & Migration

    Migration from self-managed brokers onto SQS, and honest guidance on when SQS is the wrong fit — fan-out, replay, ordering across many partitions, and complex routing generally belong on Kafka or RabbitMQ instead.

Why AceMQ

Why AWS teams add AceMQ to their SQS workloads

  • Named Engineers, Zero Cold Start

    The same senior engineers stay on your account. They know your queue layout, your redrive policies, and which consumers sit behind which queues — so a P1 call starts with diagnosis instead of orientation.

  • No Tier-1 Triage Layer

    You reach a senior engineer directly by phone, email, or Slack. Nobody collects a case description to pass along, and no support-plan tier decides how long you wait for someone who can actually read your CloudWatch metrics.

  • The Consumer Side, Which AWS Support Won't Debug

    SQS itself almost never fails. Nearly every SQS incident is a consumer problem: visibility timeout, retry behavior, concurrency, or an unhandled exception path. AWS support scopes to the service, which leaves the layer causing your incident unattended.

  • 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 oldest-message age, DLQ growth rate, empty-receive ratio, and request cost trend — so you catch a poison-message loop or a missing redrive policy before it becomes a weekend incident.

  • We Know When SQS Is the Wrong Answer

    We support RabbitMQ, Kafka, ActiveMQ, and IBM MQ at the same depth. SQS has no replay, no fan-out to multiple independent consumers, and no routing beyond one queue per destination. We will tell you plainly when you have outgrown it.

SQS estates we support

Standard and FIFO queues across AWS accounts and regions, consumers on Lambda, ECS, EKS or EC2, and SNS or EventBridge feeding queues behind VPC endpoints.

  • SQS Standard queues
  • SQS FIFO & high-throughput FIFO
  • AWS Lambda event source mappings
  • Amazon ECS & AWS Fargate
  • Amazon EKS & Kubernetes
  • EC2 & Auto Scaling groups
  • SNS, EventBridge & S3 event sources
  • Cross-account & cross-region queues
  • SSE-KMS encrypted queues
  • VPC endpoints & PrivateLink
  • AWS GovCloud & regulated accounts
  • Multi-account AWS Organizations

Version coverage

Versions we support (service and client libraries)

Amazon SQS is a managed service with no version to patch. What does have a version is the client side. We support the stack you run.

Versions we support (service and client libraries)
ComponentWhat we support
SQS queue typesStandard, FIFO and high-throughput FIFO
AWS SDKsJava, Python (boto3), .NET, Node.js and Go clients, including the retry and backoff settings of each
Amazon SQS Extended Client LibraryJava and Python (S3-backed payloads up to 2 GB, synchronous clients only, per the AWS quotas page)
Amazon SQS Java Messaging Library (JMS)JMS applications moved onto SQS
Spring Cloud AWSThe Spring listener and template layer over SQS
Lambda runtimes using event source mappingsThe runtimes you deploy, for the mapping and batching behaviour

Quota facts used on this page (AWS, SQS Developer Guide, "Quotas related to messages", read 17 November 2026): maximum message size 1,048,576 bytes (1 MiB); retention default 4 days, up to 14 days; visibility timeout default 30 seconds, up to 12 hours; batch up to 10 messages; FIFO partition limit 300 transactions per second per action in default mode.

Escalation Path

Your first hour of an Amazon SQS outage

AWS runs the queue itself; the engineer you reach works on everything that talks to it, starting from ApproximateAgeOfOldestMessage.

  1. From 15 min (by plan)first responseEmergencyProduction down: consumers stopped and the backlog is growing, SNS or EventBridge messages no longer reaching the queue, or Lambda throttled to zero
  2. 1 hourfirst responseCriticalSevere impact: duplicate processing from a short visibility timeout, a FIFO queue capped by one MessageGroupId, poison messages cycling
  3. 4 hoursfirst responseHighA DLQ filling with no alarm, redrive set too high, a bill out of line with traffic from short polling
  4. Next Dayfirst responseStandardQuestions on long polling, batching, partial batch responses, the extended client and queue and key policies

An order queue backing up at midnight and a question about batch sizes are on different clocks, set by your plan.

Want someone who already knows your queues, accounts and consumers? Request a quote or message us.

Contact Us

Pricing

What does AceMQ Amazon SQS support cost?

Amazon SQS support from AceMQ commonly costs $24,000 to $72,000 a year. The price reflects how many queues and AWS accounts you use, what consumes them (Lambda, containers, instances), and the response plan.

Prices are in US dollars (USD). Response targets begin at 15 minutes by plan, 24/7, and quote requests are answered within 1 hour. A contract that also covers RabbitMQ, Kafka or other messaging tools earns a multi-technology discount.

Support In Practice

SQS problems we have traced to their cause

Amazon SQS support and remediation work from AceMQ. Each card links to the full write-up: the queues and consumers, what AceMQ diagnosed or changed, and the result, so you can see how a similar problem was handled.

All use cases

Support options

Four tiers of Amazon SQS support.

AceMQ offers Amazon SQS support in four tiers, each handing more of the day-to-day work to AceMQ: 24/7 support and incident management for queues your team runs, proactive monitoring that catches visibility, dead-letter and scaling problems first, managed SQS inside your own AWS account, and fully managed SQS run on your behalf from AceMQ-managed AWS accounts and infrastructure, including the consumers and bridges around it. Standard and FIFO queues are covered in all four.

  • Tier 1

    Amazon SQS 24/7 Support & Incident Management

    Your developers own the queues, we take escalations

    Best for: Engineering teams using SQS in many services who want a messaging specialist when retries, backlogs or duplicates appear.

  • Tier 2

    Amazon SQS 24/7 Support with Proactive Monitoring

    Reviews of alarms, redrive and consumer settings

    Best for: Teams who find out about a full DLQ or a growing oldest-message age from a customer, not from an alarm.

  • Tier 3

    Managed Amazon SQS on Your Infrastructure

    We operate the queue layer in your AWS accounts

    Best for: Organizations that keep SQS and consumers in their own accounts but want monitoring, tuning and policy upkeep handled.

  • Tier 4

    Fully Managed Amazon SQS on AceMQ Infrastructure

    We manage producers, consumers and alarms for you

    Best for: Product teams that want dependable queue-driven workflows without dedicating staff to messaging operations on AWS.

FAQ

Amazon SQS support questions

Still have questions about Amazon SQS? Email an Expert

Who provides enterprise support for Amazon SQS?

AWS supports the SQS service under its AWS Support plans. AceMQ supports your use of it: producers, consumers, Lambda mappings, retry policy, IAM and KMS, around the clock.

What does 24/7 Amazon SQS support cover with AceMQ?

Round-the-clock incident response, a written root cause on every P1, and standing work that prevents repeats: alarms on ApproximateAgeOfOldestMessage and DLQ depth, plus visibility timeout and batching reviews.

Why is the same SQS message processed more than once?

Standard queues deliver at least once, so a duplicate is allowed. The usual causes are a handler that runs longer than the visibility timeout, a batch where one slow message pushes the others past it, and a worker that dies after doing the work but before DeleteMessage. We calibrate the timeout from measured duration, add ChangeMessageVisibility heartbeats where needed and make the consumer idempotent, because duplicates cannot be eliminated on a standard queue.

Why is our FIFO queue limited to a few hundred messages a second?

Each FIFO partition is limited to 300 transactions per second per API action in default mode, and up to 3,000 messages a second per action when you batch ten at a time. A single or low-cardinality MessageGroupId serialises everything. Spreading group IDs across real entities and enabling high-throughput mode lifts the ceiling by a large multiple, per the AWS quotas page.

How do SQS dead-letter queues and redrive work?

A redrive policy names a DLQ and a maxReceiveCount. When a message is received that many times without being deleted, SQS moves it to the DLQ. Nothing is lost, and nothing is processed either. The DLQ needs its own alarm and a retention longer than the source queue, because the original enqueue time carries over. Redrive in controlled batches after the cause is fixed.

What should we do when Lambda does not scale with the SQS backlog?

Check the event source mapping before the function: batch size, batching window, MaximumConcurrency and partial batch responses. A function that errors causes retries that look like load. Then check account and reserved concurrency. We read these settings alongside queue age so the cause is visible in one place.

Why are messages disappearing between SNS or EventBridge and our queue?

Usually an access policy. The queue policy must allow the publishing service principal, with the right source ARN condition, and an SSE-KMS queue needs a key policy that lets that service use the key. Delivery failures are logged on the SNS side, not in your consumer.

Does SQS support messages over 256 KB?

Yes. The AWS quotas page (read 17 November 2026) gives a maximum of 1 MiB per message. Beyond that, the Extended Client Library for Java and Python stores the payload in S3, up to 2 GB, and both producer and consumer must use it.

Should we be using SQS at all, or Kafka, RabbitMQ or Amazon MQ?

SQS is a good default for durable work queues with simple retry behaviour. It is not a log: there is no replay and no fan-out to independent consumer groups. If you need ordered replay or routing rules, Kafka or RabbitMQ fits better. AceMQ supports all of them, so advice is not steered by what we sell.

Can AceMQ help us secure SQS for regulated data?

Yes. We review queue and IAM policies, SSE-KMS key policies, VPC endpoints and cross-account access, and how payloads are handled. We do not certify compliance. We make sure the queue configuration matches the control you have to show.

Can we hire AceMQ for a one-off SQS review?

Yes. A review of visibility timeouts, redrive policies, alarms and queue and KMS key policies can be a single scoped job. Teams that find it useful often move to annual 24/7 support afterward, covering incidents and ongoing tuning with the same engineers.

How is Amazon SQS support priced?

Contracts run $24,000 to $72,000 a year (USD). We price by the number of queues and AWS accounts, the consumer platforms such as Lambda or ECS, how much FIFO and DLQ tuning is involved and the response plan. Tell us your setup and a quote follows within 1 hour.

Let's talk

A Queue That Never Goes Down Can Still Take You Down

Tell us your queue types, consumer platforms and AWS regions. We reply to quote requests within 1 hour.

A senior Amazon SQS engineer, not a sales queueQuote requests answered within 1 hourAd hoc engagements or an annual partnership

Production incident right now?Email an Amazon SQS Architect