Apache Pulsar support

Apache Pulsar Support for Brokers, BookKeeper and Geo-Replication

Last reviewed by Tyler Eastridge

AceMQ supports Apache Pulsar 24/7 across Americas, EMEA and APAC: self-managed clusters, StreamNative Cloud, Kubernetes and bare metal, including the BookKeeper storage layer and the metadata store. Most calls start with one of three problems: write latency spikes from bookie disks, producers throttled by backlog quotas, or bundle imbalance after a topic count change.

AceMQ supports Apache Pulsar's multi-layer architecture in production — BookKeeper bookies under disk and IO pressure causing cluster-wide write latency spikes, backlog quotas triggering unexpected producer throttling, and ZooKeeper metadata store issues cascading into broker instability. Every ticket reaches a named senior engineer who understands both the broker and bookie layers.

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

Planning a build, upgrade or migration instead? See AceMQ Apache Pulsar 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 Apache Pulsar

The Pulsar problems that reach us most, each with the pulsar-admin command, metric or ledger state we examine first.

Common Apache Pulsar production problems and how AceMQ resolves them
Symptom you're seeingWhat we check first
Write latency spikes cluster-wide with no obvious broker causeEvery publish waits for an ack quorum from bookies. We look at pulsar_storage_write_latency_* histograms on the broker and the journal and ledger metrics on the bookies. The usual cause is the journal sharing a disk with ledger writes or compaction, or a bookie with a slow disk dragging the ack quorum. Best practice is a separate fast device for the journal. We review managedLedgerDefaultEnsembleSize, managedLedgerDefaultWriteQuorum and managedLedgerDefaultAckQuorum, and bin/bookkeeper shell health commands, such as listing under-replicated ledgers.
Producers are throttled or messages are rejectedCheck backlog quotas on the namespace (pulsar-admin namespaces get-backlog-quota). The policy decides what happens at the limit: producer_request_hold, producer_exception, or consumer_backlog_eviction. A slow or stopped subscription fills the backlog and then blocks producers for every topic that quota covers. We also check publish rate limits (maxPublishRatePerTopicInMessages, maxPublishRatePerTopicInBytes), and the broker's memory and direct-memory pressure.
Retention behaves unexpectedly: messages disappear early or never expirePulsar has three separate settings: retention (how long acked messages are kept), TTL (when unacked messages are dropped from a subscription's backlog) and backlog quota. They interact. A subscription with TTL deletes backlog, and a namespace with zero retention drops acked data immediately. We read the effective policy at topic, namespace and broker level.
Brief unavailability or latency blips during normal operationBundle movement. A bundle is a hash range of topics owned by one broker. The load balancer splits and moves bundles, and topics on a moved bundle disconnect and reconnect. We check loadBalancerAutoBundleSplitEnabled and thresholds, the number of bundles per namespace, and client reconnect settings. A namespace created with too few bundles puts all hot topics on one broker.
Geo-replicated clusters drift out of syncReplication backlog per topic (pulsar-admin topics stats shows replication with backlog and delay). Common causes: network between regions, replication rate limits, a remote cluster with a quota or authorization problem, and a topic whose policy was not set on both sides. We read both clusters, not just the local one.
Broker instability that starts with metadata operationsThe metadata store handles topic creation, ownership and leader election. Slow metadata operations, large numbers of topics, or ZooKeeper session expiry cause brokers to lose ownership and thrash. We read metadata store latency and session counts, topic and subscription counts per broker, and the namespace policies that create many metadata writes.
Offloaded data cannot be read backTiered storage moves closed ledgers to S3, GCS or Azure Blob. Reads from offloaded ledgers fail if credentials, bucket policy or the offload driver settings changed, or the offload threshold (managedLedgerOffloadThresholdInBytes) and deletion lag removed local data before the offload was verified. We check offload status with pulsar-admin topics offload-status and read the broker logs.
A bookie is down and the cluster is under-replicatedBookie replacement and autorecovery. We check listunderreplicated, the auditor, and recovery rate settings so rereplication does not saturate disks while traffic runs.

Each P1 report points to the layer at fault, whether broker, bookie or metadata store, and what to change.

Producers blocked or write latency spiking across the cluster? Request a quote and we answer within an hour.

Contact Us

Deployment coverage

Which Apache Pulsar deployments does AceMQ support, and who covers what?

DeploymentWhat AceMQ coversWhat the vendor coversExcluded
Self-managed Apache Pulsar (Kubernetes, VMs, bare metal)Brokers, bookies, metadata store, proxies, Pulsar Functions, tuning, upgrades, capacityApache Pulsar is a community project with no vendor support contractHardware faults outside the agreed scope
StreamNative CloudClient and topic design, subscription behaviour, quota and policy tuning, migration planningStreamNative operates and supports its cloud; its pricing page lists 24/7 technical support in all cloud plansStreamNative platform faults
BookKeeper storage layerJournal and ledger disks, ensemble and quorum settings, autorecovery, compaction, bookie replacementCommunity projectDisk hardware
Metadata store (ZooKeeper or equivalent)Latency, sessions, sizing, backupCommunity projectNot applicable
Geo-replicationReplication policy, backlog, authentication between clusters, failover designCommunity projectNetwork provider faults
Tiered storage offloadDriver configuration, thresholds, read-back testsObject store providersProvider outages
Kubernetes and OpenShiftHelm charts, StatefulSets, storage, upgrade orderThe platform vendorControl-plane faults: see our Kubernetes support
Kafka-protocol and migration layersKoP and client compatibility, Kafka-to-Pulsar migration planningNot applicableApplications you own are reviewed, not rewritten

Support options

Who provides enterprise support for Apache Pulsar? StreamNative and independent support compared

Apache Pulsar is a community project. StreamNative, a company founded by Pulsar's original creators, sells a cloud service and support. Independent support from AceMQ covers self-managed Pulsar and the storage layer on any infrastructure.

StreamNative
AceMQ independent support
What it coversStreamNative Cloud (Serverless, Dedicated, BYOC); its pricing page also links StreamNative Private Cloud
What it coversSelf-managed Pulsar and the BookKeeper and metadata layers, plus StreamNative client-layer questions
Hours24/7 technical support listed, included in all cloud plans
Hours24/7 across Americas, EMEA and APAC
Response targetsNot specified on the page read
Response targetsResponse targets from 15 minutes, by plan
Product defectsStreamNative can fix its own distribution
Product defectsWe diagnose, gather evidence and work upstream with you
Root cause analysisPer StreamNative's terms
Root cause analysisWritten root cause analysis on every P1
PriceSet by StreamNative plan
Price$24,000 to $72,000 a year (USD), set by SLA, scope and deployment. Quote reply within 1 hour
Kafka and RabbitMQStreamNative covers Pulsar
Kafka and RabbitMQThe same team also supports Kafka and RabbitMQ at the same depth, which helps a migration decision

Dated source: StreamNative pricing page, streamnative.io/pricing, read 17 November 2026.

What's Included

Everything in your Apache Pulsar support contract

Brokers, BookKeeper bookies and the metadata store, together with geo-replication and tiered storage, covered by one contract.

  • 01

    Emergency Incident Response

    A bookie under IO pressure, a broker instability cascade, or a metadata store quorum issue. With response targets from 15 minutes, by plan, a senior engineer joins a live bridge with access across the broker and bookie layers — not a ticket acknowledgement.

  • 02

    Root Cause Analysis

    Every P1 closes with a written RCA: what failed, why, the fix applied, and the specific config or infrastructure change that prevents recurrence.

  • 03

    Broker & BookKeeper Tuning

    Journal/ledger disk separation, backlog quota and retention policy design, and bundle assignment tuning against your actual topic count and throughput profile.

  • 04

    Metadata Store Health Advisory

    Proactive monitoring of ZooKeeper (or your configured metadata store) quorum health, since a single degraded metadata node can cascade into cluster-wide broker instability.

  • 05

    Capacity & Storage Reviews

    Quarterly reviews of bookie disk headroom, backlog growth trends, and tiered storage offload behavior so you size ahead of demand rather than reacting to a throttled producer.

  • 06

    Kafka Migration & Multi-Layer Onboarding

    Migration planning from Kafka, including mapping the broker/bookie architectural split for ops teams new to it, plus geo-replication design for multi-region deployments.

Why AceMQ

Why streaming teams choose AceMQ for Pulsar

  • Named Engineers, Zero Cold Start

    The same senior engineers stay on your account. They know your bundle layout, your bookie ensemble configuration, and your replication topology — so a P1 starts with diagnosis, not twenty minutes of explaining your environment.

  • No Tier-1 Triage Layer

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

  • We Know the Storage Layer, Not Just the Broker

    Pulsar's split broker/bookie architecture is exactly where teams coming from Kafka get surprised. We treat BookKeeper as a first-class diagnostic target, not an afterthought behind the broker logs.

  • Genuine Follow-the-Sun Coverage

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

  • Proactive, Not Just Reactive

    Quarterly health checks on bookie disk and metadata store quorum health, plus shared intelligence across our support base. When a version-specific bug surfaces on one customer's cluster, every affected customer hears about it before it reaches their production.

  • Honest About the Operational Complexity

    Pulsar's multi-layer architecture buys real flexibility but real operational surface. We tell customers plainly when that trade-off fits their use case and when a simpler system would serve them better.

Pulsar environments we support

StreamNative Cloud, self-managed clusters on Kubernetes or VMs, multi-region setups, and networks with no outbound route.

  • StreamNative Cloud
  • Self-managed Pulsar clusters
  • AWS (EC2, EKS)
  • Google Cloud (GKE)
  • Microsoft Azure (AKS)
  • Kubernetes & OpenShift
  • Bare metal & on-premise
  • Tiered storage to S3/GCS
  • Geo-replicated multi-region
  • Air-gapped / no outbound access

Version coverage

Versions we support

Versions we support
Pulsar lineSupported by AceMQ
Pulsar 5.0 (LTS)Yes
Pulsar 4.0 (LTS)Yes
Other 4.x feature releases and 3.xAsk at scoping; we plan the move to a supported line

Source: Apache Pulsar release policy, pulsar.apache.org/contribute/release-policy/, read 17 November 2026. It lists 5.0 as an LTS release dated 5 October 2026 (latest listed 5.0.1) and 4.0 as an LTS release dated 21 October 2024 (latest listed 4.0.14), and defines LTS releases every 18 months and feature releases every 3 months. Components: Pulsar brokers, BookKeeper, ZooKeeper or metadata store, proxies, Pulsar Functions, Pulsar Manager, Kafka-on-Pulsar.

In practice

Sizing and design questions we answer before they become incidents

Most Pulsar P1s trace back to a sizing choice made at build time. These are the questions we put to a team in the first week of a support contract, with the reason each one matters.

What are your ensemble, write quorum and ack quorum? A common starting point is an ensemble of 3, a write quorum of 3 and an ack quorum of 2, which tolerates one slow or failed bookie without stalling writes. A larger ensemble spreads load across more bookies but widens the blast radius of a slow disk. Where a team runs an ack quorum equal to the write quorum, every write waits for the slowest bookie, and one degraded disk becomes a cluster-wide latency problem. We check the setting against the number of bookies, the failure domains (rack or zone awareness) and the latency budget.

Where does the journal live, and what shares it? The journal is a sequential, fsync-heavy write path. It wants its own fast device. Teams that put journal and ledger directories on the same volume see compaction and ledger flushes show up as publish latency. We measure fsync latency on the journal device before the first production topic.

How many bundles does each namespace have, and how many topics are hot? Topics map to bundles by hash. A namespace created with the default bundle count and a few very busy topics can put most traffic on one broker. Splitting bundles early costs little; splitting under load causes the reconnect blips we see in problem 4. We size bundles from the expected topic count and the busiest topics' rates.

Which topics need compaction, and who reads them? Compacted topics are cheap until a compaction falls behind and a consumer reading the compacted view sees stale data. We check compaction thresholds and run state per topic.

What do Pulsar Functions and connectors share with the brokers? Functions running inside the broker process share its memory and CPU. A leaking function becomes a broker incident. We prefer separate function workers for production and check their resource limits.

What happens to a subscription when its consumers go away? Durable subscriptions keep their cursor and backlog forever unless a TTL or expiry policy removes them. A forgotten subscription from a retired service is the most common source of a backlog that grows for weeks and then blocks producers. We list subscriptions with backlog and no active consumer as part of every review.

Escalation Path

Your first hour of an Apache Pulsar outage

Many Pulsar outages start in the storage layer, so the responder must look past the broker dashboards quickly.

  1. From 15 min (by plan)first responseEmergencyProducers blocked cluster-wide by a backlog quota, several bookies down with write quorum lost, or no broker owning production bundles
  2. 1 hourfirst responseCriticalWrite latency far above baseline on a key namespace, geo-replication backlog growing, offloaded data unreadable
  3. 4 hoursfirst responseHighRetention or TTL not matching policy, bundle imbalance causing blips, an under-replicated ledger after a bookie loss
  4. Next Dayfirst responseStandardEnsemble and quorum sizing, partitioned topic design, tiered storage rollout, subscription type choice

Your plan sets the clocks. A cluster-wide producer block starts the fastest one, tiered storage planning the slowest.

Want engineers who know your namespaces, bookies and replication setup? Request a quote or write to us.

Contact Us

Pricing

What does AceMQ Apache Pulsar support cost?

Apache Pulsar support is typically $24,000–72,000 a year. Price follows the number of clusters, how many brokers and bookies each has, whether geo-replication or tiered storage is on, and the response target.

Prices are in US dollars (USD). Response targets start at 15 minutes by plan, 24/7, and quote requests are answered within 1 hour. Putting Kafka or RabbitMQ under the same contract earns a multi-technology discount.

Support In Practice

Pulsar incidents we have resolved

Apache Pulsar support and remediation work from AceMQ. Each card links to the full write-up: the cluster, what AceMQ diagnosed or changed, and the result.

  • Remediation

    Apache Pulsar BookKeeper IO Remediation

    Global Payments Processor

    Resolving cluster-wide write latency caused by bookie journal and ledger device contention, and restoring write quorum headroom.

    Apache PulsarApache BookKeeperKubernetesZooKeeper
  • Consulting

    Apache Pulsar Multi-Tenancy Design

    Managed Services Provider

    Designing tenant, namespace, and policy structure so independent teams can share a Pulsar cluster without interfering with each other.

    Apache PulsarApache BookKeeperKubernetesZooKeeper
All use cases

Support options

Four tiers of Apache Pulsar support.

AceMQ offers Apache Pulsar support in four tiers, each handing more of the day-to-day work to AceMQ: 24/7 support and incident management for clusters your team runs, proactive monitoring that catches bookie, backlog and CVE problems first, managed Pulsar on your own infrastructure, and fully managed Pulsar on AceMQ infrastructure. Brokers, BookKeeper and the metadata store are covered in all four.

  • Tier 1

    Apache Pulsar 24/7 Support & Incident Management

    Your platform team runs Pulsar, we back it up

    Best for: Teams that operate brokers and bookies themselves and want a specialist for backlog, latency or replication trouble.

  • Tier 2

    Apache Pulsar 24/7 Support with Proactive Monitoring

    Reviews of bookies, quotas and bundles

    Best for: Groups that want ledger health, backlog quotas and retention policy checked before a namespace blocks producers.

  • Tier 3

    Managed Apache Pulsar on Your Infrastructure

    We operate all three layers in your estate

    Best for: Organizations keeping brokers, bookies and metadata on their own infrastructure but short on Pulsar experience.

  • Tier 4

    Fully Managed Apache Pulsar on AceMQ Infrastructure

    We run brokers, bookies and ZooKeeper for you

    Best for: Teams wanting multi-tenant streaming and queuing without staffing for brokers, BookKeeper and ZooKeeper.

FAQ

Apache Pulsar support questions

Still have questions about Apache Pulsar? Email an Expert

Who provides enterprise support for Apache Pulsar?

Pulsar is a community project and StreamNative supports its own cloud service. AceMQ covers self-managed Pulsar and BookKeeper, 24/7, with response from 15 minutes by plan. A Sev 1 with producers blocked by a backlog quota means an engineer on a bridge reading namespace policy first.

What does 24/7 Apache Pulsar support cover with AceMQ?

Round-the-clock incident response with a written root cause on every P1, plus standing work on BookKeeper health, backlog and retention, bundle balance, geo-replication and upgrades. A Sev 1 might be write quorum lost after two bookies fail.

Do you support the BookKeeper layer, or only the broker?

Both, plus the metadata store. Most Pulsar incidents have a storage-layer cause, so a support engineer who only reads broker dashboards misses it.

Why are our Pulsar producers throttled?

Usually a backlog quota. When a subscription's backlog hits the quota, the policy blocks producers or raises exceptions. We find the slow subscription, fix its consumer or TTL, and set quota policy to match the business impact.

Why do messages disappear early or never expire?

Retention, TTL and backlog quota are separate settings and interact. We read the effective policy at topic, namespace and broker level and align them.

Why is write latency high on Pulsar?

Every write waits for a BookKeeper ack quorum. Journal and ledger sharing a disk, slow bookies, compaction and ensemble settings are the usual causes.

Do you support geo-replication and tiered storage offload?

Yes. We check replication backlog on both clusters and the offload driver, thresholds and read-back.

We are migrating from Kafka to Pulsar. Can you help?

Yes. We plan topic mapping, producer and consumer changes, KoP use where it fits, and a reversible cutover. We also tell you when Kafka is the better fit; we support both.

How is Apache Pulsar different from Kafka for operations?

Pulsar splits serving from storage: brokers, BookKeeper bookies and a metadata store each fail and scale on their own. You gain flexibility but have more places to check, so incidents often start in the storage layer.

Can we hire AceMQ for a Pulsar design review only?

Yes. A review of ensemble and quorum settings, a bundle balancing fix or a Kafka-to-Pulsar mapping can stand alone as a project. Teams wanting round-the-clock cover and engineers who know their namespaces can then extend to the annual agreement afterward.

How is a Pulsar support contract priced?

Yearly contracts range from $24,000 to $72,000 (USD). We weigh cluster count, broker and bookie numbers, geo-replication, tiered storage offload, StreamNative or self-managed deployment, and the response level. Send the cluster layout and a quote comes back within 1 hour.

Let's talk

Your Pulsar Cluster's Storage Layer Shouldn't Be a Blind Spot

Tell us your cluster layout and where it runs. We reply to quote requests within 1 hour.

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

Production incident right now?Email an Apache Pulsar Architect