Configuration
Configuration that outlives one command lives in a small set of TOML
files inside one configuration directory. This is deliberately separate
from --data-dir, which holds state, not settings.
Where the configuration directory is
Lookups walk these locations in order and the first file found wins:
| Precedence | Location |
|---|---|
1. |
Exactly that directory, nothing else is searched. |
2. User config dir |
Linux: |
3. System config dir |
Linux: |
Two files live there today: cluster.toml (where the cluster’s processes
are) and notebooks.toml (the Jupyter service).
cluster.toml — the address table
waratah start reads this to find the five processes it orchestrates.
The whole file for a single-machine cluster:
compute = "127.0.0.1:7005"
[hexastore]
s = "127.0.0.1:7001"
p = "127.0.0.1:7002"
o = "127.0.0.1:7003"
[registry]
0 = "127.0.0.1:7004"
| Key | Meaning |
|---|---|
|
Address of the compute node’s internal gRPC listener. |
|
Addresses of the three storage partitions. |
|
Addresses of string-registry shards, keyed by numeric node id. |
When no cluster.toml is found, the built-in local defaults are used —
exactly the addresses above: partitions on 127.0.0.1:7001–7003,
registry on 7004, compute on 7005. A waratah start with no
configuration therefore matches a default cluster.toml byte for byte.
notebooks.toml — the Jupyter service
One flat table; unknown keys are ignored so newer engines keep reading
older files. Every value can instead be given as a waratah start flag,
and flags win: the merge order is command-line flag over notebooks.toml
value over built-in default.
| Key | Flag | Meaning |
|---|---|---|
|
|
Turn the notebook service on ( |
|
|
Root directory holding the notebooks. Default
|
|
|
Python interpreter for the sidecar. Default: a pinned virtual environment created automatically on first start. |
|
|
Loopback bind for the sidecar; |
|
|
Jupyter authentication token; minted automatically at startup when unset. |
|
|
Shut down kernels idle longer than this many seconds. Default 3600. |
What the service does with these values is covered in Jupyter Notebooks.