Emergency Incident Response
Production down, messages not flowing, or a broker that won't start. With response targets from 15 minutes, by plan, a senior engineer joins a live bridge with direct access to diagnose — not a ticket acknowledgement.
ActiveMQ support, 24/7
Last reviewed by Tyler Eastridge
AceMQ supports ActiveMQ Classic and ActiveMQ Artemis 24/7, self-managed, on Kubernetes and on Amazon MQ. Most calls start with one of three problems: producers blocked by flow control, a KahaDB journal that will not shrink, or a network of brokers that duplicates or strands messages after a split.
24/7 across Americas, EMEA and APAC, with response targets from 15 minutes, by plan, and a written root cause analysis on every P1.
Planning a build, upgrade or migration instead? See AceMQ ActiveMQ consulting and services.
Customers




Partners


Incident Triage
The broker failures behind most ActiveMQ calls, shown with the JMX attribute, config value or log we read before changing a thing.
| Symptom you're seeing | What we check first | Typical resolution |
|---|---|---|
| Producers hang with no error (Classic) | Producer flow control state; memoryLimit, storeUsage and tempUsage; whether one destination used the shared limit while a slow consumer sat behind it | A memoryLimit per destination class, a store cursor so a backlog pages to KahaDB, a sendTimeout on non-critical producers |
| KahaDB journal grows and the disk fills (Classic) | Which journal files are pinned; inactive durable subscribers in JMX holding references | Destroy abandoned durable subscribers, set an offline durable subscriber timeout, move fan-out to virtual topics |
| OutOfMemoryError at normal volume | Consumer prefetch against heap; a prefetch of 1000 on large messages across many consumers | Prefetch sized for the slowest consumer |
| Messages loop through redelivery and the DLQ grows | Redelivery policy; whether the consumer fails before or after acknowledgement; XA transaction boundaries | Correct the policy, handle poison messages, fix the acknowledgement mode |
| Duplicates or stranded messages across a network of brokers (Classic) | Network connector settings, duplex, networkTTL, conduitSubscriptions, advisory flow | Correct connector settings and TTL |
| Address full, paging, throughput collapses (Artemis) | max-size-bytes, address-full-policy, page-size-bytes against consumer drain rate; paging hides a slow consumer | Size address settings, find and fix the slow consumer |
| Clients do not reconnect to the backup | Shared-store or replication topology; file lock acquisition; whether the client URI lists the failover transport with both brokers | Correct topology and the failover URI |
| A scanner flags the broker, or the OpenWire port is reachable too widely | Which CVEs apply to your version, whether the OpenWire port is exposed beyond known clients, whether client libraries inside fat jars carry the flaw | Restrict the port first, then upgrade brokers and client libraries |
A P1 closes with a report naming the destination, subscriber or limit that caused the stall.
Deployment coverage
| Deployment | What AceMQ covers | What the vendor covers | Excluded |
|---|---|---|---|
| ActiveMQ Classic 5.x and 6.x, self-managed | 24/7 incident response, written RCA on P1s, flow control and store tuning, persistence choice (KahaDB or JDBC), CVE advice, upgrades | The Apache project publishes releases and lists which series are active; it offers community help only (activemq.apache.org, read 17 Nov 2026) | Application code and Camel routes are reviewed, not rewritten |
| ActiveMQ Artemis 2.x | Address settings, paging, journal, HA topology, upgrades | The Apache Artemis project (artemis.apache.org) | The same |
| Red Hat AMQ Broker | The Artemis behaviour underneath: addresses, paging, clients | Red Hat supports AMQ Broker for subscribers | A patched Red Hat build |
| Amazon MQ for ActiveMQ | Destination design, prefetch and memory tuning, client failover configuration | AWS runs and supports the service | AWS platform faults, broker settings the service does not expose |
| Kubernetes, OpenShift, vSphere and Tanzu | Pod and volume behaviour as it affects journals and stores, failover, rolling restarts | The platform vendor supports the platform | Control-plane faults |
| Networks of brokers, JDBC stores | Connector design, advisory behaviour, database contention on the store | The database vendor supports the database | Database engine faults |
| Protocol clients (OpenWire, AMQP, STOMP, MQTT, JMS) | Client configuration and compatibility, including Jakarta Messaging moves | None | Writing client applications |
Support options
Apache ActiveMQ is an open-source project with community help only. Commercial support comes from Red Hat for AMQ Broker, from AWS for Amazon MQ as a managed service, and from independent providers such as AceMQ for the Apache releases you run yourself.
What's Included
Classic and Artemis brokers, KahaDB or journal persistence, networks of brokers and failover pairs, covered by one agreement without per-incident fees.
Production down, messages not flowing, or a broker that won't start. With response targets from 15 minutes, by plan, a senior engineer joins a live bridge with direct access to diagnose — not a ticket acknowledgement.
Every P1 closes with a written RCA: what failed, why, the fix applied, and the specific configuration or code change that prevents recurrence. Delivered as standard, not on request.
Prefetch sizing, memory and store limits, KahaDB vs. JDBC persistence selection, and consumer concurrency tuned against your actual throughput profile and message sizes.
Proactive alerts for CVEs affecting your exact ActiveMQ version — including the 2023 CVE-2023-46604 class of RCE issues — with tested upgrade paths and mitigation guidance.
Quarterly reviews of queue depth trends, broker resource headroom, and destination growth so you size ahead of demand rather than reacting to a full store.
ActiveMQ Classic to Artemis migration planning, version upgrades with rolling procedures, and migration off ActiveMQ to RabbitMQ or Kafka where that's the right call.
Why AceMQ
The same senior engineers stay on your account. They know your broker topology, your persistence choice, and your traffic pattern — so a P1 call starts with diagnosis, not twenty minutes of you explaining your environment.
You reach a senior ActiveMQ engineer directly by phone, email, or Slack. There is no help desk collecting information to pass along, and no escalation approval process standing between you and someone who can actually fix it.
Most vendors know one. We support ActiveMQ Classic 5.x and Artemis 2.x in production, including the migration between them — which matters because the two share a name and almost nothing else architecturally.
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.
Quarterly health checks on your deployment 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.
ActiveMQ problems are often not ActiveMQ problems. We diagnose across the JVM, the OS, storage latency, network policy, and Kubernetes — because that's where the root cause frequently actually lives.
Self-managed brokers, Amazon MQ, Kubernetes, pairs sharing a store, and locked-down sites that cannot reach out to us.
Version coverage
| Series | Latest listed (activemq.apache.org / artemis.apache.org, read 17 Nov 2026) | Java | What AceMQ does on it |
|---|---|---|---|
| Classic 6.4.x | 6.4.0, 27 Oct 2026; listed as active | 17 and later | 24/7 support; the target for upgrades from 5.x |
| Classic 6.3.x | 6.3.3, 27 Oct 2026; listed as active | 17, 21, 25 | 24/7 support |
| Classic 6.2.x, 6.1.x, 6.0.x | 6.2.10 (26 Sep 2026); 6.1.8; 6.0.1; listed as inactive | 17 and later | 24/7 support and an upgrade plan to an active series |
| Classic 5.19.x | 5.19.12, 27 Oct 2026; listed as active | 11, 17, 21 | 24/7 support |
| Classic 5.18.x and earlier | 5.18.7 and earlier; listed as inactive | Per series | 24/7 support and an upgrade plan to an active series |
| Artemis 2.x | 2.57.0, 9 Sep 2026 | 17 and later | 24/7 support |
Classic 6.x moves from the javax.jms namespace to Jakarta Messaging and needs Java 17, so an upgrade from 5.x touches application code and libraries as well as the broker. We inventory clients first, then upgrade brokers and clients in an order that keeps failover working.
In practice
The two brokers need different skills. We ask which you run before anything else.
| Topic | ActiveMQ Classic | ActiveMQ Artemis |
|---|---|---|
| Persistence | KahaDB journal or a JDBC store | Append-only journal with paging to disk |
| Memory pressure | Producer flow control against memoryLimit, storeUsage, tempUsage | address-full-policy and paging |
| HA | Shared-store master and slave, or networks of brokers | Shared store or replication, with live and backup pairs |
| Native protocol | OpenWire (also AMQP, STOMP, MQTT) | Core protocol (also AMQP, STOMP, MQTT, OpenWire) |
| First place we look in an incident | Broker log and JMX for destination memory and store usage | Address and queue metrics, paging state |
Consulting
| Deliverable | What you receive |
|---|---|
| Broker health check | Memory, store and temp limits per destination, prefetch against heap, durable subscriber audit, failover test, version and CVE position |
| Performance tuning | Per-destination limits, cursor choice, prefetch sizing, persistence choice from measured throughput |
| Security review | Exposure of transport connectors, authentication and TLS on connectors, client library inventory, advisory monitoring |
| Upgrade plan | 5.x to 6.x with the Jakarta namespace move, Classic to Artemis planning, rehearsal and rollback |
| Migration planning | Moves from ActiveMQ to RabbitMQ or Kafka with protocol mapping, parallel running and canary cutover, where migration fits |
| Runbooks | Finding the slow consumer from JMX, clearing a pinned journal, failing over safely |
Escalation Path
Blocked producers do not log an error, so whoever picks up the page has to know to check flow control first.
A broker that will not start gets the fastest clock; advice on prefetch for a new consumer waits in a slower queue.
Pricing
ActiveMQ support usually costs $24,000–72,000 a year. The figure depends on whether you run Classic or Artemis, how many brokers and networks of brokers there are, KahaDB or JDBC persistence, and the response target you need.
Prices are in US dollars (USD). Response targets start at 15 minutes by plan, 24/7, and quote requests are answered within 1 hour. Adding Artemis or Classic brokers, Kafka or other messaging to one contract earns a multi-technology discount.
Support In Practice
ActiveMQ support and migration work from AceMQ. Each card links to the full write-up: the broker estate, what AceMQ diagnosed or changed, and the result.
Regional Payments Processor
Four live sessions to find why every producer on an ActiveMQ Classic 5.16 pair stalled when one settlement consumer slowed, and to give each destination its own memory limit.
Healthcare Benefits Administrator
A broker's data volume kept filling although its queues were nearly empty. The cause was a durable topic subscription for an application retired a year earlier.
Industrial Equipment Manufacturer
A manufacturer with ActiveMQ Classic brokers at plants and in the cloud had no inventory of versions or clients when CVE-2023-46604 landed. AceMQ contained it, found what was exposed, and patched brokers and clients.
Support options
AceMQ offers ActiveMQ support in four tiers, each handing more of the day-to-day work to AceMQ: 24/7 support and incident management for brokers your team runs, proactive monitoring that catches store, memory and CVE problems first, managed ActiveMQ on your own infrastructure, and fully managed Artemis on AceMQ infrastructure. Classic 5.x, 6.x and Artemis 2.x are covered in all four.
Your team owns the brokers, we escalate
Best for: Application teams running Classic or Artemis who need a broker specialist when flow control or failover goes wrong.
Regular broker audits and failover tests
Best for: Groups that want durable subscribers, store usage and exposed connectors checked before they cause an outage.
We operate your brokers in your environment
Best for: Companies whose brokers must stay on their own hosts or private network, but whose team is stretched thin.
We run your ActiveMQ brokers as a service
Best for: Teams who rely on JMS messaging but would rather not run the brokers, stores and failover themselves.
One partner for ActiveMQ
Our engineers also handle RabbitMQ and Kafka, Camel routes, and the planning when a move off ActiveMQ makes sense.
The Apache project offers community help only. Red Hat supports AMQ Broker, its own product built on ActiveMQ Artemis, and AWS supports Amazon MQ as a managed service. Independent providers such as AceMQ support the Apache releases themselves, Classic and Artemis, on premises, in the cloud or on Kubernetes, 24/7, with response targets from 15 minutes, by plan.
Round-the-clock response from 15 minutes on the top plans. For a Sev 1 where producers are blocked, an engineer joins a bridge, reads flow control state and store usage, and you receive a written root cause. Standing work includes subscriber audits, failover tests and upgrade planning.
Yes. They share a name and have different architectures, persistence models and configuration. We support Classic 5.x and 6.x and Artemis 2.x in production, and plan the migration between them.
Almost always producer flow control. A destination hit its memory limit, usually because a consumer is slow, and the broker throttles senders. In Classic we check memoryLimit, store and temp usage and prefetch, then add per-destination limits and a store cursor.
A journal file is deleted only when no message in it is still needed, and one retained message can pin a whole file. The usual holder is an inactive durable subscriber or one unacknowledged message. We find it from JMX and fix the cause.
Yes. Amazon MQ removes some operations, but flow control, prefetch, destination design and client failover remain yours, and they cause most incidents.
Plan it. Apache lists 6.4.x, 6.3.x and 5.19.x as active series. Classic 6.x needs Java 17 and uses Jakarta Messaging, so the upgrade touches application code. ActiveMQ has had serious flaws, including the widely exploited OpenWire remote code execution issue CVE-2023-46604, so we restrict exposure, put mitigations in place and support the version you run while the upgrade is prepared.
Classic speaks OpenWire natively plus AMQP, STOMP and MQTT; Artemis adds its core protocol. Both implement JMS. Support covers these, Camel routes, JDBC or KahaDB persistence, networks of brokers and Java client configuration.
Yes, and we say whether you should. We plan migrations to RabbitMQ or Kafka with protocol mapping, parallel running and canary cutover, and we often advise that a fixable ActiveMQ is cheaper than a migration.
We read memory, store and temp usage per destination, prefetch against heap, the durable subscriber list, failover behavior, exposed connectors, and your version's known CVEs. The result is a ranked plan with the first fixes marked.
Red Hat and AWS consult on their own products, while AceMQ consults on Apache ActiveMQ itself, Classic and Artemis. The consultants also cover the 24/7 rota, so their designs account for how brokers actually fail.
Yes. We support legacy ActiveMQ versions that no longer receive upstream patches, including bug workarounds and backported mitigations where official fixes are unavailable. Many regulated environments cannot upgrade on the community's timeline, and that constraint is a normal part of enterprise support rather than a disqualifier. For security fixes on ActiveMQ Classic 5.15 to 5.18 and on Artemis, AceMQ's OSSeva platform backports CVE patches, so an estate can stay on its version while the upgrade is planned.
No. A one-off broker health check, a stuck-journal cleanup or a 5.x to 6.x upgrade plan can be bought alone. If you later want 24/7 cover and named engineers who know your destinations and clients, the annual option continues from that work.
Contracts range from $24,000 to $72,000 a year (USD). We weigh Classic against Artemis, the number of brokers and environments, whether they run as a single node, a failover pair or a network of brokers, and your response level. Share the topology and a quote follows within 1 hour.
Let's talk
Send us your ActiveMQ version, Classic or Artemis, and your broker topology. We reply to quote requests within 1 hour.
Production incident right now?Email an ActiveMQ Architect