Jobs

The jobs store keeps:

Each session’s background jobs, the ones cmd & starts, and each job’s output.
The processes each instance of the workspace runs, which ps lists.
When each instance was last seen, so the others can tell it is gone.

`type`	Keeps them	Keys
`ram`	In the process’s memory. Jobs end with the process.
`redis`	On Redis, under `key_prefix` and the workspace id. A job’s output stays a day after its last write.	`url`, `key_prefix`, default `mirage:`

The group key is store.jobs. A job’s output is a stream other processes follow, so disk is refused here, and a disk store keeps its jobs in memory.

On Redis

YAML
Python
TypeScript

workspace_id: agent
store:
  type: disk
  root: /var/lib/mirage
  jobs:
    type: redis
    url: redis://redis.internal:6379/0

from mirage import Workspace
from mirage.workspace.store import DiskWorkspaceStateStore, RedisWorkspaceStateStore

store = DiskWorkspaceStateStore(
    root="/var/lib/mirage",
    jobs=RedisWorkspaceStateStore(url="redis://redis.internal:6379/0"),
)
ws = Workspace(mounts, workspace_id="agent", store=store, owns_store=True)

import { DiskWorkspaceStateStore, RedisWorkspaceStateStore, Workspace } from '@struktoai/mirage-node'

const store = new DiskWorkspaceStateStore({
  root: '/var/lib/mirage',
  jobs: new RedisWorkspaceStateStore({ url: 'redis://redis.internal:6379/0' }),
})
const ws = new Workspace(mounts, { workspaceId: 'agent', store, ownsStore: true })

Here the rest of the state stays on the machine’s disk, and every process on the machine that builds the workspace sees the same jobs.

Run several instances of one workspace

Every process that builds the workspace under the same id on one Redis is an instance of it, on one machine or several, behind a load balancer or as replicas of one agent.

YAML
Python
TypeScript
Servers

workspace.yaml

workspace_id: agent
store:
  type: redis
  url: redis://redis.internal:6379/0
mounts:
  /data:
    vfs: s3
    config:
      bucket: agent-data

from mirage import MountMode, Workspace
from mirage.vfs.s3 import S3VFS, S3Config
from mirage.workspace.store import RedisWorkspaceStateStore

ws = Workspace(
    {"/data": S3VFS(S3Config(bucket="agent-data"))},
    mode=MountMode.WRITE,
    workspace_id="agent",
    store=RedisWorkspaceStateStore(url="redis://redis.internal:6379/0"),
    owns_store=True,
)

import { MountMode, RedisWorkspaceStateStore, S3VFS, Workspace } from '@struktoai/mirage-node'

const ws = new Workspace(
  { '/data': new S3VFS({ bucket: 'agent-data' }) },
  {
    mode: MountMode.WRITE,
    workspaceId: 'agent',
    store: new RedisWorkspaceStateStore({ url: 'redis://redis.internal:6379/0' }),
    ownsStore: true,
  },
)

for server in http://mirage-a:8765 http://mirage-b:8765; do
  curl "$server/v1/workspaces" \
    -H "Authorization: Bearer $MIRAGE_AUTH_TOKEN" \
    -H 'Content-Type: application/json' \
    --data-binary @workspace.json
done

On the servers, workspace.json holds {"id": "agent", "config": {...}} with the same config on each. The instances then share:

Each session’s background jobs: jobs, wait, kill %N and fg %N work on any instance, whichever started the job.
Processes: ps lists the other instances’ as owner:pid, and kill owner:pid stops one.
Cleanup: an instance counts as gone 30 seconds after it was last seen, and the next one to start stops the programs it left running.
Sessions, history and symlinks, since the whole store is on Redis.

A session sees only its own processes, on every instance; a profile with processes: workspace widens that to every session’s. Shell describes the details.

Getting Started

Agents

VFS

Integration

CLIs

Mirage Shell

Agent Profiles

Policy

Runtimes

Infrastructure

Configuration

Notifications

Help

On Redis

Run several instances of one workspace

​On Redis

​Run several instances of one workspace

On Redis

Run several instances of one workspace