Sync¶
Cross-cluster job state synchronization.
Overview¶
The SyncManager class enables sharing job states across multiple compute clusters via git. This is useful when:
- Running experiments across multiple clusters
- Sharing job status with collaborators
- Maintaining a central record of all job states
Classes¶
SyncManager¶
SyncManager
¶
SyncManager(sync_dir: Optional[Union[str, Path]] = None, config: Optional[Config] = None)
Manager for cross-cluster job state synchronization.
Handles writing sync files and optionally committing/pushing them to git for sharing across clusters.
Attributes:
| Name | Type | Description |
|---|---|---|
sync_dir |
Directory where sync files are stored. |
|
hostname |
Current machine's hostname. |
Example
sync_mgr = SyncManager() sync_mgr.sync_all() # Refresh and write sync file sync_mgr.push() # Git commit and push
Initialize the sync manager.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
sync_dir
|
Optional[Union[str, Path]]
|
Directory for sync files. If None, uses config. |
None
|
config
|
Optional[Config]
|
Configuration object. If None, uses global config. |
None
|
Source code in src/slurmkit/sync.py
sync_collection
¶
Sync a single collection's job states.
Refreshes job states from SLURM and returns the updated data.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
name
|
str
|
Collection name. |
required |
Returns:
| Type | Description |
|---|---|
Dict[str, Any]
|
Collection sync data dictionary. |
Source code in src/slurmkit/sync.py
sync_all
¶
Sync all collections and write sync file.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
collection_names
|
Optional[List[str]]
|
List of collection names to sync. If None, syncs all collections. |
None
|
Returns:
| Type | Description |
|---|---|
Dict[str, Any]
|
Complete sync data dictionary. |
Source code in src/slurmkit/sync.py
push
¶
Git commit and push sync files.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
message
|
Optional[str]
|
Commit message. If None, uses a default message. |
None
|
Returns:
| Type | Description |
|---|---|
bool
|
True if successful, False otherwise. |
Source code in src/slurmkit/sync.py
get_sync_file_path
¶
Get the path to a sync file for a hostname.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
hostname
|
Optional[str]
|
Hostname for the sync file. Defaults to current host. |
None
|
Returns:
| Type | Description |
|---|---|
Path
|
Path to the sync file. |
Source code in src/slurmkit/sync.py
Usage Example¶
from slurmkit.sync import SyncManager
from slurmkit import CollectionManager
# Initialize sync manager
sync = SyncManager()
# Get a collection
manager = CollectionManager()
collection = manager.get("my_experiment")
# Sync collection state to git-tracked file
sync.sync_collection(collection)
# Push changes to remote
sync.push(message="Update job states from cluster-a")
CLI Usage¶
The sync functionality is typically used via the CLI: