Nexus Repository Requirements: Sonatype's Sizing Profiles, the H2 Limit, and the Monthly Cost

Sonatype's system requirements for Nexus Repository are written for enterprises, which is why a team that wants a Maven and Docker proxy reads them and wonders whether it needs a 16 GiB server. This guide goes through the published figures, checked on 1 November 2026, explains which apply to a small deployment, and prices three configurations of the Nexus Repository package on MassiveGRID PaaS, built on Virtuozzo Application Platform (formerly Jelastic).

What Sonatype publishes

The requirements page says plainly that there is not a single defined requirement, because of the range of use cases, and gives sizing profiles instead. The smallest, Extra-Small, is 4 vCPU and 16 GiB per node on one node; Small is 8 vCPU and 32 GiB; Medium the same per node on three nodes; Large 32 vCPU and 128 GiB on three; Extra-Large 48 vCPU and 192 GiB on four. Highly available deployments must meet the profile on every node. The memory rule inside a node is to allocate up to two thirds of the RAM to Nexus Repository, leaving a third for the system and its buffers; the page defers heap settings to a separate memory overview topic.

Performance, the page notes, is bounded primarily by IO, disk and network, rather than CPU. Disk has a hard rule: at least 4 GB of free disk at all times, below which the database switches to read-only. The application directory was about 390 MB as of August 2024; the data directory is the one that grows, and formats such as Docker and Maven can use very large amounts of storage, 500 GB easily. For the database, the embedded H2 supports up to 200,000 requests a day or 100,000 components, container-based deployments are not supported on H2, and Sonatype recommends an external PostgreSQL for all production deployments.

Reading the profiles for a team

The XS profile describes a Nexus that proxies Maven Central, npm and Docker Hub for a company and hosts its releases, with the H2 database and a load that stays under the 200,000-requests-a-day line. A ten-developer team with a CI pipeline on every merge is nowhere near that line on an ordinary day and some way past the 4 GB free-disk rule within a month of caching Docker images, which is the point: for a team, disk is the requirement and memory is the comfort.

So the practical floor sits below XS. An 8 GiB node with two thirds, about 5 GiB, for Nexus runs a team's proxy and private repositories well; XS is the figure for a company whose builds depend on Nexus being fast under concurrent CI load; above that, PostgreSQL and a profile with more IO.

The three configurations on PaaS and their cost

The Nexus Repository package on MassiveGRID PaaS installs Nexus Repository OSS with persistent blob storage, an HTTPS web UI and repository endpoints, and an initial administrator. Cloudlets are 128 MiB RAM plus 400 MHz CPU at $0.003372 per cloudlet-hour, $2.46 per cloudlet-month, no licence fee for the OSS edition.

Two things make the real bill lower than the first column. Billing is for the cloudlets a container uses each hour, and a Nexus that serves a team in one time zone idles at a fraction of its limit overnight; and blob storage, billed per gigabyte, is the line that actually tracks usage, so a cleanup policy on cached Docker layers is worth more than a smaller node. The PaaS pricing calculator prices other sizes and schedules.

Disk, the requirement that bites first

Sonatype's two storage statements, 4 GB free at all times and 500 GB easily for Docker and Maven, describe the same risk from both ends. A Nexus that proxies Docker Hub caches every layer any build pulls, and a team that builds daily fills tens of gigabytes a month without noticing until the database goes read-only. The fix is policy, not disk: a cleanup policy on proxy repositories that removes components not downloaded in thirty days, a scheduled compaction task, and a disk alert on the platform's load-alerts add-on at 80%.

On the platform, persistent storage grows from the dashboard without downtime, which turns the 4 GB rule from an outage into a line on the invoice; but the per-gigabyte billing means the cleanup policy still decides what Nexus costs.

H2 or PostgreSQL, and when to move

The OSS package starts on H2, which Sonatype supports up to 200,000 requests a day or 100,000 components and does not support for container-based deployments. For a team, that is years of headroom; for a company, the request count is reached long before the component count. When it is, the platform's PostgreSQL primary-secondary cluster package is the external database Sonatype recommends, on the internal network, with the TLS add-on if a review asks for encryption in transit, and Nexus's migration tooling moves the metadata.

The Nexus on PaaS post sets up the proxy, the hosted repositories and the CI configuration that makes builds reproducible, and the add-on checklist covers the backups a repository that holds release artifacts needs.

Frequently Asked Questions

What is the minimum server for Nexus Repository?

Sonatype publishes profiles rather than a minimum; the smallest, XS, is 4 vCPU and 16 GiB on one node, with up to two thirds of the RAM allocated to Nexus, as checked on 1 November 2026. A team's proxy runs well below that, which is why the Starter configuration on MassiveGRID PaaS allocates 8 GiB.

How much disk does Nexus need?

Sonatype requires at least 4 GB of free disk at all times, below which the database becomes read-only, and warns that Docker and Maven formats can easily use 500 GB. On the platform, storage grows from the dashboard and is billed per gigabyte, so a cleanup policy on proxied repositories is the real cost control.

Can I run Nexus on the embedded H2 database?

Yes, up to 200,000 requests a day or 100,000 components, and not for container-based deployments; Sonatype recommends an external PostgreSQL for production. The platform's PostgreSQL cluster package provides that database on the internal network when the limit approaches.