---
title: "Manage clusters: the settings that matter most"
description: "A broker starts with no settings, and most installations change very few. The ones that matter most, some read only at the first start, and where each one is explained."
---

> Queen MQ documentation, for AI agents
> Complete self-contained summary of Queen MQ: https://queenmq.com/llms-brief.txt
> Fetch that first when the question is about the product rather than about this page.
> Index of all pages: https://queenmq.com/llms.txt

# Manage clusters: the settings that matter most

A broker starts with no settings at all, and most installations change very few. These are the ones that matter most. Some are read only once, when a node creates its data directory, so choose them before the first start. The pages of this section explain each one where it is used, and [Configuration](/reference/configuration/) lists every setting.

**Read only at the first start**

| Setting | Default | What it does |
| --- | --- | --- |
| `QUEEN_QLOG_SHARDS` | 0 | `0`: each queue has its own log files. `K`: all queues share K logs, so a write syncs at most K files. For thousands of queues. |
| `QUEEN_RAFT_PEERS` | unset | The nodes of a cluster (see [Run a cluster with Docker](/manage/docker/)). After the first start, members change only through the membership API. |
| `QUEEN_RAFT_DIR` | /var/lib/queen/raft | The data directory, one per node. |

**How fast old data leaves**

| Setting | Default | What it does |
| --- | --- | --- |
| `PARTITION_CLEANUP_SECONDS` | 30 days | An empty partition is deleted after this long without a write. At least 60. |
| `QUEEN_RAFT_TXN_WINDOW_MIN_S` | 900 | The least time a message's dedup record is kept. Nothing is deleted sooner, partitions included. |
| `QUEEN_RAFT_RETENTION_SCAN_PER_S` | 50000 | Partitions the retention scan visits a second. On 2.6.0, set `40000` when partitions come and go fast (see [Many short-lived partitions](/manage/configuration/#many-short-lived-partitions)). |
| `QUEEN_RAFT_TRACE_RETENTION_S` | 7 days | How long message traces are kept. |

**Limits**

| Setting | Default | What it does |
| --- | --- | --- |
| `QUEEN_RAFT_DISK_HIGH_PCT` | 85 | Above this share of the disk, writes that would grow it answer 507. |
| `QUEEN_EPHEMERAL_MAX_BYTES` | 256 MiB | The memory for ephemeral queues. Above it, their pushes answer 503. |

**Security**

| Setting | Default | What it does |
| --- | --- | --- |
| `QUEEN_RAFT_TOKEN` | unset | The secret the nodes of a cluster share. |
| `JWT_ENABLED` | false | Check a token on every request. |
| `QUEEN_ENCRYPTION_KEY` | unset | The key of encrypted queues: 64 hex characters, the same on every node. |

**Upgrades and logs**

| Setting | Default | What it does |
| --- | --- | --- |
| `QUEEN_RAFT_CLUSTER_VERSION_MS` | 1000 | `0` keeps the way back to the previous release open. |
| `LOG_LEVEL`, `QUEEN_LOG_JSON` | info, false | How much the node logs, and whether each line is a JSON object for a log collector. |

**Why:** most settings have a default that is right for most installations. The few above decide the shape of a cluster, how fast old data leaves, and who may talk to it, so they are worth a decision before the first start.

## On these pages

| Page | What it covers |
| --- | --- |
| [Cluster with Docker](/manage/docker/) | Three nodes with Docker Compose, and what happens when one stops. |
| [Kubernetes](/manage/kubernetes/) | A StatefulSet with Helm or the operator, probes, and rolling updates. |
| [Configuration by need](/manage/configuration/) | The settings for each need, from trying QueenMQ to surviving a site. |
| [Upgrades and monitoring](/manage/upgrades-monitoring/) | One node at a time, the way back, and what to watch. |
| [Recovery](/manage/recovery/) | What to do when a node, a disk or a majority is gone. |
| [Backups](/manage/backups/), [Standby cluster](/manage/standby/) | Copies of the data, and a second cluster to fail over to. |
| [Security](/manage/security/), [Tenants and API keys](/manage/tenants/) | Who may talk to the cluster, and many tenants on one. |
| [Kubernetes operator](/manage/operator/), [Terraform](/manage/terraform/) | Clusters and queues declared as resources. |

Source: https://queenmq.com/manage/index.mdx