How-to
This page lists every procedure in this documentation by task, not by the chapter that contains the procedure. Chapters are sorted by the question they answer, so the steps for one task are in one place and never appear twice. Use this page when you know what you want to do but not which chapter covers the task.
The explanations are on the linked pages. Each row names a task and points to the page that has the steps for that task, the costs, and the reading that shows that the work is done.
A new cluster runs from packages to a verified mount
| What you want to do | Where it is |
|---|---|
| Check that the machines, the network, and the plan are ready | Pre-flight checklist |
| Prepare a machine: repository, disks, name resolution, RAM, CPU | Prepare the operating system |
| Install the Master Server | Install server components |
| Install a Chunkserver and give it its disks | Install server components |
| Install a Metalogger | Install server components |
| Install the GUI and the command line interface | Install server components |
| Install the license on a new cluster PRO (PRO Edition only) | Install a license |
| Replace the license on a running cluster PRO (PRO Edition only) | Install a license |
| Install the client on Linux, FreeBSD, or macOS | Install clients |
| Install the Windows Client | Windows Client |
| Mount the file system for the first time and confirm it works | First mount and verification |
| Make a mount survive a reboot | First mount and verification |
| Start the cluster in the right order | Starting and stopping |
| Stop the cluster in the right order | Starting and stopping |
| Run more than one component, or more than one process of one, on a machine | Several processes on one machine |
| Build a cluster that spans sites PRO (PRO Edition only) | Divide a new cluster into locations |
| Copy an existing data set into the cluster | Migrating data in |
Everyday work on a mounted share needs no administrator rights
| What you want to do | Where it is |
|---|---|
| Recover a deleted file | Using the file system |
| Decide how long deleted files stay recoverable | Using the file system |
| Take a snapshot of a file or a directory tree | Using the file system |
| Find out where a file is stored and whether it is whole | Using the file system |
| Measure how much a directory holds | Using the file system |
| Read a file's Storage Class and its extra attributes | Using the file system |
Access, placement, and space are policy that you can change on a running cluster
| What you want to do | Where it is |
|---|---|
| Admit a machine to the file system, or take its access away | Access control |
| Put a password on an export | Access control |
| Map the identity a remote user acts with | Access control |
| Give an administrator machine the rights to change the cluster | Access control |
| Make a file undeletable, append-only, or immutable | Access control |
| Cap how much one mount may ask of the Master Server | Access control |
| Make the server modules prove who they are | Access control |
| Keep the GUI away from people who should not reach it | Access control |
| Label the Chunkservers so placement can tell them apart | Data placement tasks |
| Define a Storage Class of your own | Data placement tasks |
| Put a file or a directory under a Storage Class | Data placement tasks |
| Move data from fast disks to slow ones as it ages | Data placement tasks |
| Keep older data as erasure-coded parts instead of copies | Data placement tasks |
| Assign a class by file name instead of by hand | Data placement tasks |
| Check where the data actually lies | Data placement tasks |
| Read how much space the cluster and one subtree hold | Space management |
| Hold a subtree inside a share of the cluster | Space management |
| Free space that deleted files still occupy | Space management |
| Change the options a mount is running with | Client and mount tasks |
| Give a machine a subtree instead of the whole file system | Client and mount tasks |
| Hand a mount the password its export demands | Client and mount tasks |
| Unmount without losing the last writes | Client and mount tasks |
| Clear the session a mount left behind | Client and mount tasks |
| Change the shared code the server modules register with | Access control |
| Change the password on an export | Access control |
Hardware changes are planned work with an order to keep
| What you want to do | Where it is |
|---|---|
| Take a Chunkserver down for a reboot or a short update | Disks and Chunkservers |
| Take a Chunkserver down for hardware service or a move | Disks and Chunkservers |
| Service several Chunkservers in one window | Disks and Chunkservers |
| Add a Chunkserver to a running cluster | Disks and Chunkservers |
| Add disks to a Chunkserver | Disks and Chunkservers |
| Retire a disk that still works | Disks and Chunkservers |
| Replace a disk that still works | Disks and Chunkservers |
| Remove a Chunkserver from the cluster for good | Disks and Chunkservers |
| Replace a disk that has already failed | Chunkserver and disk recovery |
The Master Server's tasks are about metadata, and in PRO about the group
| What you want to do | Where it is |
|---|---|
| Force the metadata to be written to disk now | Master Server operations |
| Take a copy of the metadata off the machine | Master Server operations |
| Read a metadata dump without starting a Master Server | Master Server operations |
| Back up a directory incrementally between two snapshots PRO (PRO Edition only) | Master Server operations |
| Apply a configuration change to a running Master Server | Master Server operations |
| See the state of every Master Server PRO (PRO Edition only) | Master Server operations |
| Move the Leader role to a machine you choose PRO (PRO Edition only) | Master Server operations |
| Add a Master Server to a group that is already serving PRO (PRO Edition only) | Master Server operations |
| Take a Master Server out of the group PRO (PRO Edition only) | Master Server operations |
| Move a Master Server to different hardware PRO (PRO Edition only) | Master Server operations |
| Prove that a metadata copy restores | Master Server operations |
| Move a running cluster to different addresses | Addresses and ports |
| Move a service to another port | Addresses and ports |
| Turn a label-separated cluster into locations PRO (PRO Edition only) | Location tasks |
| Add a site to a cluster that already has locations PRO (PRO Edition only) | Location tasks |
| Move a module from one location to another PRO (PRO Edition only) | Location tasks |
| Take a whole site out of service on purpose PRO (PRO Edition only) | Location tasks |
| Delete a location PRO (PRO Edition only) | Location tasks |
| Serve a file in the cluster as a block device | Block device operations |
| Give a mapped block device a file system, resize it, or unmap it | Block device operations |
An upgrade runs one way, so the way back is decided first
| What you want to do | Where it is |
|---|---|
| Find out whether two versions may run in one cluster | Version compatibility |
| Check that the cluster is fit to be upgraded | Pre-upgrade checks |
| Upgrade the components in the right order | Order of upgrade |
| Upgrade the Master Servers without a pause in service PRO (PRO Edition only) | Rolling upgrade |
| Go back to the version you came from | Rollback |
| Move an instance from Community Edition to PRO | CE and PRO |
| Move an instance from PRO back to Community Edition | CE and PRO |
| Read what changed before moving to 5.x | What changed in 5.x |
Another system reaches MooseFS through a gateway
| What you want to do | Where it is |
|---|---|
| Serve the file system over SMB | SMB |
| Serve the file system over NFS | NFS |
| Serve the file system over FTP | FTP |
| Give Kubernetes pods volumes on MooseFS | MooseFS CSI driver |
| Hand Kubernetes pods a mount made on each node | Node mount and hostPath |
| Store Proxmox virtual machine images | Proxmox |
| Store XCP-ng virtual machine images | XCP-ng |
| Reach the file system from a program without mounting it | mfsio API |
A symptom that you can see starts in Troubleshooting
Repair work is not listed here, because it is not sorted by task. You find repair work by what you see, not by what you want to do. Start from the symptom is the table that matches a symptom to the page that covers the symptom. What to watch lists the readings that a running cluster needs and how often to read them.