Research · Original data

Managed service limits index: what Amazon MQ, MSK, ElastiCache, Neon and Supabase let you change

Every limit in this index comes from the vendor's own documentation, with a link and the date it was checked. It is refreshed quarterly.

  • Checked 9 Nov 2026
  • Refreshed quarterly
  • Sources linked

This managed service limits index records 13 documented limits across six services from four vendors: Amazon MQ for RabbitMQ, Amazon MSK and Amazon ElastiCache from AWS, then Neon, Supabase and Instaclustr. Examples are a three-node RabbitMQ cluster, 30 brokers per MSK cluster on ZooKeeper, and Supabase Free projects paused after one week. Each row links its source and was checked on 9 Nov 2026.

The headline finding

Turnkey services are built for the average workload, and their own documentation says where they stop. Amazon MQ for RabbitMQ documents a three-node cluster, Amazon MSK documents 30 brokers on ZooKeeper and 60 on KRaft, Neon documents compute ceilings by plan, and Supabase pauses Free projects after one week of inactivity.

13documented limits recorded
6services covered, from 4 vendors
9 Nov 2026date every source was checked

A documented limit is not a defect. It matters when a workload sits at the edge of it, and most workloads do not.

Method: sources, dates and what was checked

Each row states what the vendor's documentation says and no more. Every limit links the page it was read from and shows the date it was checked.

Sources.
AWS documentation pages for Amazon MQ, Amazon MSK and Amazon ElastiCache, and the pricing or documentation pages of Neon, Supabase and Instaclustr. No third-party summaries.
Checked.
Each page was read on 9 Nov 2026. A row shows the date it was read, and the date moves only when the page is read again.
Wording.
Limits are paraphrased from the page, and numbers are as the page gives them.
Scope.
Limits that are unsourced, or that could not be confirmed on the page, are left out. No pricing, performance or uptime figures are recorded here.
RDS and Atlas.
No Amazon RDS or MongoDB Atlas row is published until a limit is recorded with a source and a date.

The data table

The table lists every recorded limit with its source and the date checked. It is reproduced below as CSV so it can be copied.

ServiceAreaDocumented limitSourceChecked
Amazon MQ for RabbitMQCluster sizeA cluster is three nodes, one per availability zone, or a single instance.AWS docs9 Nov 2026
Amazon MQ for RabbitMQSettings and operator accessThe configurable-values table lists five broker values: consumer timeout, heartbeat, operator policy changes, relaxed quorum queue redeclaration checks and HTTP security headers. AWS documents further options for authentication and resource limits.AWS docs9 Nov 2026
Amazon MQ for RabbitMQPluginsA documented list of 11 plugins. Static shovels are not supported, a shovel cannot target a private broker, and federation cannot use an upstream queue or exchange in a private broker. The delayed-message exchange and stream plugins are not on the list.AWS docs9 Nov 2026
Amazon MQ for RabbitMQUpgradesAWS manages upgrades to the latest patch version in the maintenance window, and moves a broker to the next minor version when its current minor version is retired.AWS docs9 Nov 2026
Amazon MQ for RabbitMQPer-node ceilingsResource limits can be set only up to documented maximums for each instance type, for example 5,000 connections and 15,000 channels per node on mq.m7g.large.AWS docs9 Nov 2026
Amazon MSKCluster sizeThe documented maximum is 30 brokers per cluster on ZooKeeper and 60 on KRaft. To update the configuration or the Kafka version of a cluster, the number of partitions per broker must first be under the documented limits.AWS docs9 Nov 2026
Amazon ElastiCache (Redis OSS)Admin commandsFor clusters running Redis OSS, ElastiCache lists these commands as unavailable: bgrewriteaof, bgsave, config, debug, migrate, replicaof, save, slaveof, shutdown and sync.AWS docs9 Nov 2026
NeonCompute ceilingUp to 2 CU (8 GB RAM) on Free, 16 CU (64 GB RAM) on Launch, and on Scale 16 CU autoscaling or a fixed size up to 56 CU (224 GB RAM).Neon pricing9 Nov 2026
NeonWhere it runsEight AWS regions. New projects can no longer be created in Azure regions, and a project's region is fixed when it is created.Neon regions docs9 Nov 2026
NeonIdle projectsFree-plan projects inactive for 90 days or more are subject to deletion as of 5 October 2026.Neon regions docs9 Nov 2026
SupabaseFree planA 500 MB database; projects are paused after 1 week of inactivity; a limit of 2 active projects.Supabase pricing9 Nov 2026
SupabaseDisk included on Pro8 GB of disk per project is included on Pro; more disk is billed at $0.125 per GB.Supabase pricing9 Nov 2026
InstaclustrPricing scope (BYOC)Under bring-your-own-cloud, the price covers only Instaclustr's management; infrastructure is paid directly to the cloud provider.Instaclustr pricing9 Nov 2026

Source: the vendor's own documentation or pricing page, linked in each row, with the date it was read. Re-check the page before relying on a figure; vendors change their documentation.

managed-service-limits.csv
service,area,documented_limit,source_url,checked
"Amazon MQ for RabbitMQ","Cluster size","A cluster is three nodes, one per availability zone, or a single instance.","https://docs.aws.amazon.com/amazon-mq/latest/developer-guide/rabbitmq-broker-architecture.html","9 Nov 2026"
"Amazon MQ for RabbitMQ","Settings and operator access","The configurable-values table lists five broker values: consumer timeout, heartbeat, operator policy changes, relaxed quorum queue redeclaration checks and HTTP security headers. AWS documents further options for authentication and resource limits.","https://docs.aws.amazon.com/amazon-mq/latest/developer-guide/configurable-values.html","9 Nov 2026"
"Amazon MQ for RabbitMQ","Plugins","A documented list of 11 plugins. Static shovels are not supported, a shovel cannot target a private broker, and federation cannot use an upstream queue or exchange in a private broker. The delayed-message exchange and stream plugins are not on the list.","https://docs.aws.amazon.com/amazon-mq/latest/developer-guide/rabbitmq-basic-elements-plugins.html","9 Nov 2026"
"Amazon MQ for RabbitMQ","Upgrades","AWS manages upgrades to the latest patch version in the maintenance window, and moves a broker to the next minor version when its current minor version is retired.","https://docs.aws.amazon.com/amazon-mq/latest/developer-guide/upgrading-brokers.html","9 Nov 2026"
"Amazon MQ for RabbitMQ","Per-node ceilings","Resource limits can be set only up to documented maximums for each instance type, for example 5,000 connections and 15,000 channels per node on mq.m7g.large.","https://docs.aws.amazon.com/amazon-mq/latest/developer-guide/rabbitmq-resource-hard-limit.html","9 Nov 2026"
"Amazon MSK","Cluster size","The documented maximum is 30 brokers per cluster on ZooKeeper and 60 on KRaft. To update the configuration or the Kafka version of a cluster, the number of partitions per broker must first be under the documented limits.","https://docs.aws.amazon.com/msk/latest/developerguide/limits.html","9 Nov 2026"
"Amazon ElastiCache (Redis OSS)","Admin commands","For clusters running Redis OSS, ElastiCache lists these commands as unavailable: bgrewriteaof, bgsave, config, debug, migrate, replicaof, save, slaveof, shutdown and sync.","https://docs.aws.amazon.com/AmazonElastiCache/latest/dg/ClientConfig.RestrictedCommands.html","9 Nov 2026"
"Neon","Compute ceiling","Up to 2 CU (8 GB RAM) on Free, 16 CU (64 GB RAM) on Launch, and on Scale 16 CU autoscaling or a fixed size up to 56 CU (224 GB RAM).","https://neon.com/pricing","9 Nov 2026"
"Neon","Where it runs","Eight AWS regions. New projects can no longer be created in Azure regions, and a project's region is fixed when it is created.","https://neon.com/docs/introduction/regions","9 Nov 2026"
"Neon","Idle projects","Free-plan projects inactive for 90 days or more are subject to deletion as of 5 October 2026.","https://neon.com/docs/introduction/regions","9 Nov 2026"
"Supabase","Free plan","A 500 MB database; projects are paused after 1 week of inactivity; a limit of 2 active projects.","https://supabase.com/pricing","9 Nov 2026"
"Supabase","Disk included on Pro","8 GB of disk per project is included on Pro; more disk is billed at $0.125 per GB.","https://supabase.com/pricing","9 Nov 2026"
"Instaclustr","Pricing scope (BYOC)","Under bring-your-own-cloud, the price covers only Instaclustr's management; infrastructure is paid directly to the cloud provider.","https://www.instaclustr.com/pricing/","9 Nov 2026"

The same data is published as a CSV file, one row per limit with its source link and the date checked.

What it means for buyers

The limits matter in three situations: a cluster that needs more nodes than the plan allows, a workload that depends on a setting or plugin the service blocks, and an estate that spans more than one cloud.

Scope the claim.
These are limits of turnkey services. Specialists differ: CloudAMQP, for example, sells five-node dedicated plans and installs custom plugins on request (CloudAMQP, Oct 2026).
Check the edge before the move.
If a workload fits well inside every row, a turnkey service is usually the cheaper choice.
Read the upgrade rows as calendar questions.
Teams that hold a validated version through a change freeze need to know who decides when a version moves.

Amazon MQ and Amazon MSK are AWS services (Amazon MQ docs, Amazon MSK docs, checked 9 Nov 2026), so a multi-cloud estate needs a different service in each cloud.

Where OSStend fits

OSStend is managed open source, tailored to your workload. An engineer designs the cluster around the measured load, so the plan ceilings in the table above do not apply.

How that works is described on how it works, and what a move involves is in the Amazon MQ to RabbitMQ migration guide. Each service in the index has an alternative page: Amazon MQ, Amazon MSK, Amazon ElastiCache, Neon, Supabase and Instaclustr. OSStend is operated by AceMQ, with 160+ customers, 10+ years and 24/7 on call.

Contact us if

  • Your workload is at or past one of the limits above.
  • You need the same design in more than one cloud or in your data center.
Talk to an engineer

Not a fit if

  • Your workload fits comfortably inside a standard plan.
  • You want a free tier or a card sign-up.

Update log

The index is re-checked every quarter. Each update re-reads every source, moves the date, and records any limit that was added, changed or removed.

DateChange
9 Nov 2026Re-read every source. Added three Neon rows, two Supabase rows and one Instaclustr row, each with its own source and date. Reworded the Amazon MQ settings, plugins, upgrades and per-node rows, the Amazon MSK row and the Amazon ElastiCache row to match what each AWS page states.
1 Nov 2026Published the dataset as a downloadable CSV file. No limit was added, changed or removed.
30 Oct 2026First publication: seven limits across Amazon MQ for RabbitMQ, Amazon MSK and Amazon ElastiCache.
1 Feb 2027Next review: every row re-checked against its source, with the date updated and any change noted here.
When sourcedRows for Amazon RDS and MongoDB Atlas, once a limit is recorded with a source and a date.
Talk to an engineer

Talk to an engineer about your limits

Bring the service you run and the limit you are near. An engineer will review it with you and sketch the design in 45 minutes.

  • A design sketch for your workload, not a plan menu
  • A zero-downtime cutover path for what you run today
  • A price for that design, after the review
See pricing →

Your request goes to the OSStend team. No newsletter sign-up.

Get a quoteTalk to an engineer
2Engines
3Where should it run?
4What do you need?

Your request goes to the OSStend team. No newsletter sign-up.