> ## Documentation Index > Fetch the complete documentation index at: https://artie.com/docs/llms.txt > Use this file to discover all available pages before exploring further. # Create a Managed Kafka cluster > Create an Artie Managed Kafka cluster from the Artie Dashboard: pick a name, region, and size, choose which IP ranges can connect, and start the brokers. ## What you'll need * To be a company admin, signed in with a work email. * A card on file, if your company doesn't pay Artie yet. You can add one during setup. * The IP ranges your apps connect from, in CIDR form, for example `203.0.113.4/32`. ## Setup **1. Create a new cluster** In the Artie Dashboard, go to **Kafka**, click **Add a cluster**, and choose **Create a new cluster**. Setup saves as you go, so you can leave and come back to it. **2. Name and size** | Setting | What to enter | | - | - | | **Name** | 3 to 100 characters. You can rename the cluster later. | | **Region** | **US East (N. Virginia)** only for now. More regions are coming. | | **Brokers** | 3 to 30. Defaults to 3. | | **Instance type** | Memory optimized (r8g, r8i) or general purpose (m8g, m8i). Defaults to `r8g.large`. Graviton (`g`) types run on Arm; Intel (`i`) types run on x86. | | **Storage per broker** | 20, 50, 80, or 100 to 1,024 GiB. Defaults to 100 GiB. | **Tiered storage** is on by default. Kafka keeps each message for its topic's retention. With tiered storage, only the newest messages stay on broker disks, and every message is also copied to S3, which costs far less than disk. Consumers read recent messages from disk at full speed and older ones from S3, a little slower. Click **Customize** to change: * **Keep on broker disks**: how many days of messages stay on broker disks, from 1 to 14. Defaults to 3. Older messages are removed from the disks but stay in S3 until the topic's retention ends. For example, with 3 days on disk and 7 days of retention, the last 3 days are on disk and in S3, and days 4 to 7 are only in S3. * **Disk cap per partition**: an optional size limit, in GiB. Once a partition holds this much on disk, Kafka removes its oldest messages from the disk, even if they're newer than the days above. They stay in S3. Leave it empty for no cap. **Storage autoscaling** is on by default. It grows every broker's disk before it fills and emails your admins each time. Disks never shrink. You can change: * **Grow when a disk reaches**: when the fullest broker's disk is this full, Artie grows every broker's disk. Defaults to 65%. * **Grow until disks are about**: how full the fullest disk is after it grows, which sets how much room each growth adds. Defaults to 55%, and must be below the first value. For example, a 100 GiB disk holding 65 GiB grows to about 120 GiB, so it's about 55% full. * **Never grow past**: the largest a broker's disk can get, in GiB. Defaults to no limit. **3. Network access** **Restrict access by IP** is on by default. Under **Allowed IP ranges**, enter one IPv4 CIDR per line, or separate them with commas. Artie rejects IPv6 ranges, ranges without a `/` size, `/0`, ranges that don't start at their first address, and duplicates. > [!WARNING] > > If you turn off **Restrict access by IP**, any client with the password can connect from anywhere. **4. Review and create** Check the cluster, network access, and estimated monthly cost, then click **Create cluster**. ![The Review and create step, listing the cluster's name, region, brokers, instance type, storage, tiered storage, storage autoscaling, and allowed IP ranges](https://mintcdn.com/artie/UZcMDJIwww2OXKYR/assets/managed-kafka/review-and-create.png?fit=max\&auto=format\&n=UZcMDJIwww2OXKYR\&q=85\&s=1ea2aa478ef7dfaf5f94ab4d743933fe) If your company doesn't pay Artie yet, click **Add a card** first. Artie needs a card, not a bank account, before it starts the cluster. ## After you create the cluster Starting the brokers usually takes 6 to 7 minutes. You can leave the page. The cluster moves through these steps: 1. Creating the cluster's resources 2. Issuing the TLS certificate 3. Starting a machine for each broker 4. Starting the Kafka brokers 5. Ready When it's ready, [connect to your cluster](https://artie.com/docs/managed-kafka/connect.md). ## Advanced > [accordion] What if the cluster doesn't deploy? > > If a step fails, or the cluster is still starting after 20 minutes, click **Try again** on the setup page or **Retry** on the cluster's page. Our team is notified when a deploy fails. If that doesn't fix it, click **Talk to us** or **Contact us**. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.