NodeLink
A NodeLink is a folder that several nodes share. Every node you attach it to mounts it at the same path, so what one agent writes there, agents on the other nodes can read: a build cache, a shared dataset, or the sign-ins of your model accounts. Share creates one for exactly that last case, so an account signed in on one node is signed in on all of them.
A NodeLink's files live on a storage. Jaah-managed storage needs no setup when its nodes all run on machines Jaah runs in one location; otherwise the storage is a file system in your own cloud, reached through a cloud connection, and several NodeLinks can share one storage. A node reaches that file system over a private route to its mount target, so a node outside the file system's network needs one you provide, such as peering or a VPN.
The chip beside a NodeLink's name shows its lifecycle: pending until its storage is ready, provisioning while Jaah creates it, then ready, when nodes can mount it, or failed; deleting while it is removed. Once it is ready, each attached node reports its own mount as Mounted, Not mounted with the reason, or another state.
Everything in a NodeLink is readable and writable by every process on every node it is attached to, so attach it only to nodes you trust with all of its files. NodeLink is shown only while your plan includes it.
- How NodeLink works — where the files live, a NodeLink's lifecycle, how a node reaches them, what each node reports, and deleting a NodeLink and its storage.
- Attach a NodeLink to nodes — create a NodeLink, give it storage, and attach or detach nodes.
- Set up the network for NodeLink — what to configure when your nodes run in the file system's VPC, in another VPC, on premises or in another cloud, with or without a public IP.
- Share model-account sign-ins between nodes — set up the NodeLink that lets every attached node reuse one sign-in.
- Connect a cloud account — add the cloud connection that cloud nodes launch through and NodeLink storage uses.
- Delete a NodeLink and its storage — unmount and delete a NodeLink, then delete or keep the storage behind it.
- Pin the image a connection launches — boot every new cloud node from one image, and re-pin it when due.
- Repair a NodeLink's file-system policy — when nodes aren't in the file-system policy, or it reads Outdated, have Jaah rewrite it.
- Put a NodeLink on a file system you already have — use an existing AWS EFS, its mount target and access point.
- Share files between tasks on different nodes — give a project's tasks one folder, whichever node they run on, and tell its agents what it is for.
- Fix a NodeLink a node can't mount — read why on the node's NodeLinks tab, fix it by its reason, or detach it.
- Change where a NodeLink mounts — change its mount path, and check every attached node mounts it there.
- Keep a NodeLink's files in GCP or Azure — put a NodeLink on a new or existing GCP Filestore or Azure Files (NFS) share.
- Change the AWS credential a NodeLink uses — switch its EFS storage to another credential, then check Jaah still reads the file system.
- Find which NodeLink uses a file system — start from a file system id in your cloud and find its NodeLinks and the nodes that mount them.
Every field of a NodeLink and its tabs is described in the Networks reference.