Google Cloud Pub/Sub support

Google Cloud Pub/Sub Support for Pull, Push and Ordered Subscriptions

Last reviewed by Tyler Eastridge

AceMQ supports Google Cloud Pub/Sub 24/7 across Americas, EMEA and APAC: pull, StreamingPull, push, BigQuery and Cloud Storage subscriptions, with the subscribers on GKE, Cloud Run, Dataflow and Compute Engine. Most calls start with one of three problems: messages redelivered because the ack deadline lapses, ordering keys stalling throughput, or a push backlog growing without errors.

Pub/Sub is managed, but the failure modes are still yours: ack deadlines expiring under slow handlers, ordering keys serializing your throughput to one stream, push subscriptions backing up behind a 503, and exactly-once semantics quietly degrading. Every ticket reaches a named senior engineer who already knows your subscription topology.

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

Planning a build, upgrade or migration instead? See AceMQ Google Cloud Pub/Sub 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 Google Cloud Pub/Sub

Redelivery loops, ordering-key bottlenecks, quiet push backlogs and vanishing messages: each symptom with the metric or log we read before touching code.

Common Google Cloud Pub/Sub production problems and how AceMQ resolves them
Symptom you're seeingWhat we check first
The same messages arrive again and again and the backlog never drainsRising expired_ack_deadlines_count is the signature. The subscriber takes messages, does not ack them in time, and Pub/Sub redelivers. Causes: an ack deadline shorter than handler time; client-side flow control that lets the library hold more messages than the handler can finish, so the lease extension runs out; a subscriber that crashes after processing but before ack; or a pull consumer that acks after slow downstream I/O. We set the deadline and client maxOutstandingMessages and maxOutstandingBytes from measured handler time, keep the extension period sensible (the exactly-once documentation notes a maximum of 600 seconds for minDurationPerAckExtension), and make handlers idempotent.
Throughput collapsed after enabling message orderingOrdering is per ordering key, and each ordering key is limited to 1 MBps (Google Cloud quotas page). A single constant ordering key serialises the topic and caps it at that rate; messages also wait behind a failed one in the same key until it is acked or the key is resumed after a publish error. We redesign the key to the entity that needs order, check the publisher's resumePublish handling after errors, and confirm subscribers are in the same region.
Push subscription backlog climbs and the endpoint logs show no errorsPush delivery is flow-controlled by the endpoint's responses and speed. Slow responses, 4xx and 5xx, or responses past the ack deadline throttle delivery and the backlog grows. We read push_request_latencies and response codes, check authentication (OIDC token audience mismatch is a common 401), and compare the endpoint's concurrency with Pub/Sub's push rate. For high volume, a pull or StreamingPull subscriber is often more controllable.
Exactly-once subscription still produces duplicatesPer Google's documentation, exactly-once delivery works only on pull subscriptions (including StreamingPull), only within a region (a subscriber in another region may see duplicates), and only the latest ack ID is valid. Without the ack-with-response API the client cannot know whether an ack failed, and the message returns. Publisher retries that create separate publishes get separate message IDs, so they are separate messages. We check client library versions, the ack-with-response path and the publisher's retry design.
Messages silently disappear before consumers see themCheck message retention (7 days default, up to 31 days with topic retention, per the quotas page), subscription expiry (31 days of inactivity by default), subscription filters that drop non-matching messages by design, and a dead-letter policy forwarding after the maximum delivery attempts.
Publisher latency spikes and publish futures time outPublisher batching settings (maxMessages, maxBytes, maxLatency), request size limits (10 MB and 1,000 messages per publish request, per the quotas page), network egress from a VPC with restricted perimeters, and retry settings in the client. We trace publish latency by region and client.
The dead-letter topic fills with messages that look fineA dead-letter policy forwards a message after a set number of delivery attempts (the default is 5). A transient downstream outage, a deadline too short, or an application bug causes attempts without a real poison message. The Pub/Sub service account needs publisher permission on the dead-letter topic and subscriber permission on the source subscription, or forwarding fails silently and messages loop. We audit IAM, then write the replay procedure.
Subscribers behind VPC Service Controls or private networksPerimeter rules, Private Google Access and egress rules block publishes or pulls with error messages that point away from the perimeter. We read the audit logs for the denial reason.

Each P1 report points to the subscription setting, client flow control or IAM binding that caused it, and the change we made.

Subscription backlog climbing and subscribers stalled? Ask for a quote now; we answer within the hour.

Contact Us

Deployment coverage

Which Google Cloud Pub/Sub setups does AceMQ support, and who covers what?

SetupWhat AceMQ coversWhat Google coversExcluded
Pull and StreamingPull subscriptionsAck deadline, flow control, concurrency, idempotency, exactly-once behaviour, client library settingsThe Pub/Sub service, its SLA and quotasYour application code is reviewed, not rewritten
Push subscriptionsEndpoint behaviour, authentication, retry policy, backlog analysisThe service's push delivery engineYour endpoint's infrastructure faults
Ordering keysKey design, publisher resume handling, throughput limitsService limits (1 MBps per ordering key)Ordering guarantees not designed into the key
BigQuery and Cloud Storage subscriptionsSchema mapping, failure and dead-letter behaviour, cost reviewGoogle runs the export subscriptionBigQuery or storage faults
Dead-letter topics and retry policiesDelivery attempts, IAM bindings, replay proceduresService behaviourBusiness decisions about which messages to discard
GKE, Cloud Run, Cloud Functions, Dataflow, Compute Engine subscribersScaling on backlog, graceful shutdown, concurrency, connection reuseThe Google Cloud compute platformsGKE control-plane faults: see our Kubernetes support
Multi-cloud and hybrid (AWS, Azure, on-premises brokers, Kafka)Bridges, connectors and migration pathsGoogle side onlyOther vendors' services
VPC Service Controls and restricted perimetersPerimeter rules, private access and audit-log analysisGoogle Cloud networking platformYour corporate network

Support options

Who provides enterprise support for Google Cloud Pub/Sub? Google Cloud Support and independent support compared

Google supports Pub/Sub as a service under Google Cloud Customer Care. It does not review your subscriber code, ack handling or topic design. Independent support from AceMQ covers that layer, and works with whichever Google support package you hold.

Google Cloud support
AceMQ independent support
What it coversThe Pub/Sub service across Google Cloud: availability, quotas, platform faults
What it coversHow publishers and subscribers use Pub/Sub, plus design and cost
PackagesBasic (all customers), Enhanced and Premium unlock extra services such as case escalation and follow-the-sun requests
PackagesOne contract; plan set by SLA, scope and deployment
HoursP1 cases default to follow-the-sun handling; 24/7 availability depends on service and language
Hours24/7 across Americas, EMEA and APAC
Response timesSet in Google's Technical Support Services Guidelines:
Response timesResponse targets from 15 minutes, by plan
Root cause analysisPlatform-level
Root cause analysisWritten root cause analysis on every P1
PriceSet by Google Cloud
Price$24,000 to $72,000 a year (USD), set by SLA, scope and deployment. Quote reply within 1 hour
Beyond GoogleGoogle Cloud products
Beyond GoogleThe same engineers support Kafka, RabbitMQ and other brokers, which helps in hybrid and migration work

Dated source: Google Cloud support procedures documentation, docs.cloud.google.com/support/docs/procedures, read 17 November 2026. Short note: raise a Google case when the service itself misbehaves. Raise one with us when the symptom is in subscriber behaviour, ordering, redelivery or dead-letter flow.

What's Included

Everything in your Google Cloud Pub/Sub support contract

Subscribers, ack handling, ordering, push endpoints, dead-letter flow and cost are covered across your projects, since Google runs the service and you own its use.

  • 01

    Emergency Incident Response

    Backlog growing, subscribers stalled, or a publish path failing. With response targets from 15 minutes, by plan, a senior engineer joins a live bridge with access to diagnose against your project — not a ticket acknowledgement.

  • 02

    Root Cause Analysis

    Every P1 closes with a written RCA: what failed, why, the fix applied, and the specific subscription or client change that prevents recurrence. Delivered as standard, not on request.

  • 03

    Throughput & Latency Tuning

    Ack deadlines, flow control limits, ordering-key design, batch settings, and subscriber concurrency tuned against your real message sizes and handler timings rather than library defaults.

  • 04

    Cost & Quota Review

    Pub/Sub bills on data volume, retention, and inter-region traffic. We find the subscriptions retaining seven days for no reason, the cross-region pulls nobody intended, and the retry loops paying twice for the same message.

  • 05

    Hybrid & Multi-Cloud Integration

    Pub/Sub rarely lives alone. We support the bridges into Kafka, RabbitMQ, and on-prem MQ estates — Dataflow paths, Pub/Sub connectors, and custom relays — including ordering and delivery guarantees across the boundary.

  • 06

    Migration & Design Review

    Kafka or RabbitMQ to Pub/Sub, Pub/Sub Lite consolidation, and topic and subscription design reviews before you scale a schema decision you cannot easily reverse.

Why AceMQ

Why Pub/Sub teams want one senior engineer

  • Named Engineers, Zero Cold Start

    The same senior engineers stay on your account. They know your topic layout, your ordering strategy, and your subscriber runtime — so a P1 call starts with diagnosis, not twenty minutes of you explaining your project structure.

  • No Tier-1 Triage Layer

    You reach a senior messaging engineer directly by phone, email, or Slack. No help desk collecting information to pass along, no case number that sits overnight, and no escalation approval standing between you and someone who can fix it.

  • We Know What Pub/Sub Replaced

    Most Pub/Sub incidents come from patterns carried over from Kafka, RabbitMQ, or IBM MQ that Pub/Sub models differently. We support all of them, so we recognize a consumer-group assumption or a queue-depth mental model on sight.

  • 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.

  • Beyond What Google Support Covers

    Google will tell you the service is healthy, and it usually is. The problem is normally your ack deadline, your ordering keys, or your subscriber's flow control — the application-side surface a platform vendor cannot debug for you.

  • Full-Stack, Not Just the Broker

    Pub/Sub problems are often not Pub/Sub problems. We diagnose across the client library, the JVM or runtime, GKE resource limits, VPC egress, and IAM — because that's where the root cause frequently actually lives.

Pub/Sub setups we support

GKE, Cloud Run and Compute Engine subscribers, Dataflow pipelines, BigQuery and Cloud Storage subscriptions, Pub/Sub Lite, and perimeters under VPC Service Controls.

  • Google Cloud Platform (all regions)
  • GKE & GKE Autopilot
  • Cloud Run & Cloud Functions
  • Compute Engine
  • Pub/Sub Lite
  • Dataflow & Apache Beam pipelines
  • BigQuery & Cloud Storage subscriptions
  • Multi-cloud (GCP + AWS or Azure)
  • Hybrid with on-premise brokers
  • VPC Service Controls / restricted perimeters
  • Java, Python, Go, Node.js, and .NET clients
  • Terraform-managed topics & subscriptions

Version coverage

Versions we support

Pub/Sub is managed, so there is no broker version. What we support is the feature set and client stack you use.

Versions we support
AreaWhat we support
Subscription typesPull, StreamingPull, push, BigQuery, Cloud Storage
Delivery featuresOrdering keys, exactly-once delivery (pull only), dead-letter topics, retry policies, subscription filters, message retention
Client librariesJava, Python, Go, Node.js and .NET Pub/Sub client libraries
Related servicesDataflow and Apache Beam, Cloud Run, Cloud Functions, GKE

Facts used (Google Cloud, Pub/Sub quotas and exactly-once documentation, read 17 November 2026): message data up to 10 MB; publish request up to 10 MB and 1,000 messages; 1 MBps per ordering key; 10,000 subscriptions attached to a topic; 7-day default retention and up to 31 days with topic retention; exactly-once delivery supported on pull subscriptions only, within a region.

In practice

Choosing between pull, StreamingPull and push, from what breaks in production

The subscription type decides which failures you will see, so it is worth choosing from the failure modes, not from the tutorial.

StreamingPull keeps a long-lived stream open per subscriber and is the default in the high-level client libraries. It gives the best throughput and latency, and the client library handles lease extension. Its failure modes are client-side: too much outstanding data held in memory, streams reset by proxies, and lease extension that runs out. When messages come back repeatedly, we read the client's flow-control settings first.

Pull (unary) gives the most control: you decide when to request and when to ack. It suits batch-style consumers and workers that scale from the backlog. Failure modes are polling too slowly, a pool of workers each holding a batch, and acking after a long downstream call.

Push hands delivery to an HTTPS endpoint, which suits Cloud Run and functions. The service controls the rate based on your endpoint's responses, so endpoint latency becomes delivery rate. Failure modes are slow responses, authentication mismatch on the OIDC token audience, and cold starts that miss the ack deadline. Push subscriptions do not support exactly-once delivery (Google Cloud documentation, read 17 November 2026).

BigQuery and Cloud Storage subscriptions write messages straight to a table or bucket with no subscriber code. Failure modes are schema mismatch (messages that do not fit the table end in the dead-letter path or are retried), quota on the destination, and cost.

Our rule of thumb: use StreamingPull for sustained services, pull for scheduled or autoscaled workers, push when the consumer is already a request-driven service with predictable latency, and a direct export subscription when no transformation is needed. We review each subscription against the traffic shape and its handler, and say plainly when moving a subscription type would remove a class of incidents.

Escalation Path

Your first hour of a Google Cloud Pub/Sub outage

A backlog that is growing against a deadline is urgent, a quota question is not. Each severity below carries a clock from your plan.

  1. From 15 min (by plan)first responseEmergencyProduction down: publishes failing, every subscriber stalled, or a backlog growing on a business-critical subscription
  2. 1 hourfirst responseCriticalSevere degradation: endless redelivery from expired ack deadlines, an ordering key holding a stream, push endpoint throttling delivery
  3. 4 hoursfirst responseHighDead-letter topic filling, duplicates despite exactly-once, VPC Service Controls blocking pulls or publishes
  4. Next Dayfirst responseStandardQuestions on retention, subscription filters, publisher batching, quota limits and migration to or from Pub/Sub

Your plan sets the clock for each severity, so a draining backlog is answered before a retention question.

Need someone who already understands your topics, subscriptions and ack handling? Request a quote or write to us.

Contact Us

Pricing

What does AceMQ Google Cloud Pub/Sub support cost?

Google Cloud Pub/Sub support usually runs $24,000–72,000 a year. Topic and project counts, subscription types, the platforms your subscribers run on and the response plan you select set the quote.

Prices are in US dollars (USD). Response targets start at 15 minutes by plan, 24/7, and quote requests are answered within 1 hour. Covering other messaging tools in the same contract can earn a multi-technology discount.

Support In Practice

Pub/Sub cases we have closed

Google Cloud Pub/Sub support and remediation work from AceMQ. Each card links to the full write-up: the topics and subscribers, what AceMQ diagnosed or changed, and the result, so you can see how a similar problem was handled.

  • Support

    Stopping a Pub/Sub Redelivery Spiral on GKE Subscribers

    Digital Wallet Provider

    Subscribers on GKE fell behind, Pub/Sub redelivered messages they were still working on, and members received duplicate notifications. AceMQ found the gap between the ack deadline and the handler.

    Google Cloud Pub/SubKubernetesJavaPostgreSQL
  • Support

    Restoring Pub/Sub Throughput After Message Ordering Was Enabled

    Logistics Technology Provider

    Throughput fell to a trickle after message ordering was switched on. AceMQ found a low-cardinality ordering key and publishers that had stopped sending for a key after a single failed publish.

    Google Cloud Pub/SubKubernetesPython
  • Remediation

    Pub/Sub Dead-Letter Topics That Never Received a Message

    Telecom Billing Provider

    Six subscriptions had a dead-letter policy, and poison messages kept cycling anyway. AceMQ found missing service agent permissions and a dead-letter topic with nothing subscribed to it.

    Google Cloud Pub/SubTerraformPython
All use cases

Support options

Four tiers of Google Cloud Pub/Sub support.

AceMQ offers Google Cloud Pub/Sub support in four tiers, each handing more of the day-to-day work to AceMQ: 24/7 support and incident management for subscriptions your team runs, proactive monitoring that catches ack deadline, ordering and dead-letter problems first, managed Pub/Sub inside your own Google Cloud project, and fully managed Pub/Sub run on your behalf from AceMQ-managed Google Cloud projects and infrastructure, including the subscribers and bridges around it. Pull and push subscriptions are covered in all four.

  • Tier 1

    Google Cloud Pub/Sub 24/7 Support & Incident Management

    Your teams own publishers and subscribers, we escalate

    Best for: Teams that publish and subscribe at scale and want a messaging specialist when ack or ordering behavior goes wrong.

  • Tier 2

    Google Cloud Pub/Sub 24/7 Support with Proactive Monitoring

    We review subscriptions, quotas and cost regularly

    Best for: Teams who want retention, dead-letter alarms and cross-region pulls examined before the invoice or an outage exposes them.

  • Tier 3

    Managed Google Cloud Pub/Sub on Your Infrastructure

    We operate the messaging layer in your projects

    Best for: Organizations that keep topics, subscribers and IAM inside their own Google Cloud projects but want specialists in charge.

  • Tier 4

    Fully Managed Google Cloud Pub/Sub on AceMQ Infrastructure

    We run your Pub/Sub messaging for you

    Best for: Product teams that want events flowing reliably while someone else watches backlogs, deadlines and dead-letter queues.

FAQ

Google Cloud Pub/Sub support questions

Still have questions about Google Cloud Pub/Sub? Email an Expert

Who provides enterprise support for Google Cloud Pub/Sub?

Google supports the service under its Cloud support plans. AceMQ supports how you use it, 24/7, with response from 15 minutes on top plans. A Sev 1 example: a subscription backlog that never drains because ack deadlines expire, where an engineer checks expired_ack_deadlines_count against handler time on a bridge. P2 is 1 hour.

Pub/Sub is fully managed by Google. What is there to support?

The service rarely fails. The incidents sit in how it is used: ack deadlines, flow control, ordering keys, push endpoints, dead-letter IAM and publisher batching. Those are your code and settings, outside what a platform support plan reviews.

Why are our Pub/Sub messages delivered again and again?

Because they are not acked before the deadline. Check expired_ack_deadlines_count, handler duration against the ack deadline, and client flow control. We set both from measured handler time and make handlers idempotent.

Why did throughput drop after we enabled ordering?

Each ordering key is limited to 1 MBps, and messages behind a failed one wait. One constant key serialises the topic. Use a key per entity and handle resumePublish after errors.

Does exactly-once delivery stop duplicates?

Only in part. It applies to pull subscriptions, within a region, and requires the client to use ack with response. Publisher retries that produce separate publishes create separate messages. Handlers should still be idempotent.

Why is our push subscription backlog growing?

Push delivery is throttled by slow or failing endpoint responses. We read push latency and response codes, check authentication, and review endpoint concurrency. Pull is an option when you need control.

How do dead-letter topics work on Pub/Sub?

After the configured maximum delivery attempts (default 5), Pub/Sub forwards the message to the dead-letter topic. The service account needs the right IAM roles on the topic and subscription. We audit those, alarm on dead-letter count and write a replay procedure.

Do you support Pub/Sub alongside Kafka or on-premises brokers?

Yes. We support hybrid designs and bridges between Pub/Sub and Kafka, RabbitMQ or IBM MQ, and plan migrations in either direction.

What limits of Pub/Sub cause production problems?

Ordering key throughput (1 MBps per key), publish request size (10 MB and 1,000 messages), 7-day default retention, subscription expiry after inactivity, and regional scope of exactly-once. We map each limit to your design.

Can you help us migrate to or from Pub/Sub?

Yes, with reversible cutover patterns such as dual publishing and bridged subscriptions, and verification that ordering and delivery semantics still hold.

Is there a way to try AceMQ on one Pub/Sub issue?

Yes. A subscription design review, a dead-letter IAM audit or a Kafka-to-Pub/Sub migration plan can be quoted as its own engagement. If you then want 24/7 response and engineers who already know your topics, the work converts to the annual agreement without starting over.

How do you price Pub/Sub support?

Contracts typically cost $24,000 to $72,000 a year (USD). We look at how many topics and projects you run, pull, push or StreamingPull subscriptions, subscriber platforms such as GKE or Cloud Run, and the response plan. This covers our work, not Google's usage charges. Quotes come back within 1 hour.

Let's talk

Your Pub/Sub Backlog Shouldn't Be Your First Warning

Tell us your subscription types and subscriber platforms. We reply to quote requests within 1 hour.

A senior Google Cloud Pub/Sub engineer, not a sales queueQuote requests answered within 1 hourAd hoc engagements or an annual partnership

Production incident right now?Email a Google Cloud Pub/Sub Architect