small fixes

.github/workflows/integration_tests.yml

@@ -94,6 +94,7 @@ jobs:
         run: ./.github/workflows/stop-goth.sh
 
       - name: Prepare artifact name
+        if: always()
         run: |
           echo ARTIFACT_NAME=logs-example-${{ inputs.BRANCH }}-${{ matrix.example_name }} | sed "s|/|_|g" >> $GITHUB_ENV
 

golem-cluster.full.yaml

@@ -1,6 +1,6 @@
 # This is Ray on Golem Full yaml 
 # - the example config for testnet cluster, 
-# - with all the properties potentially changable by the user
+# - with all the properties potentially changeable by the user
 
 # Ray on Golem cluster name
 cluster_name: golem-cluster
@@ -74,7 +74,7 @@ provider:
         max_env_per_hour_price: 0.5
 
         # If not provided payment will be sent only after cluster is stopped. 
-        # Required for long running clusters
+        # Required for long-running clusters
         payment_interval_hours:
           minimal: 12
 
@@ -87,7 +87,7 @@ available_node_types:
     # The maximum number of worker nodes of this type to launch
     max_workers: 0
 
-    # The node type's CPU and GPU resources - leave it empty for autodetection
+    # The node type's CPU and GPU resources - leave it empty for auto-detection
     resources: {}
 
     # Additional parameters specific for this node type added on top of node_config from provider.parameters.node_config
@@ -146,4 +146,3 @@ worker_start_ray_commands: [
 # This will be removed by ray-on-golem on the fly.
 head_node: True
 worker_nodes: True
-

golem-cluster.yaml

@@ -18,7 +18,6 @@ provider:
   use_internal_ips: true
   module: "ray_on_golem.provider.node_provider.GolemNodeProvider"
   parameters:
-
     # Blockchain used for payments.
     # `holesky` means running free nodes on testnet,
     # `polygon` is for mainnet operations.

ray_on_golem/server/cluster/nodes.py

@@ -43,9 +43,9 @@ class ClusterNode(WarningMessagesMixin, NodeData):
     _cluster: "Cluster"
     _golem_service: GolemService
     _manager_stack: ManagerStack
+
     _sidecars: Collection[ClusterNodeSidecar]
     _ssh_public_key_data: str
-
     _activity: Optional[Activity] = None
     _start_task: Optional[asyncio.Task] = None
     _warning_messages: List[str] = None
@@ -293,7 +293,8 @@ async def verify_ssh_connection(
             "uptime",
         ]
 
-        retry = CLUSTER_MONITOR_RETRY_COUNT
+        num_retries = CLUSTER_MONITOR_RETRY_COUNT
+        retry = num_retries
         while True:
             try:
                 await run_subprocess_output(*ssh_command_parts)

ray_on_golem/server/cluster/sidecars.py

@@ -89,6 +89,8 @@ def __init__(self, node: "ClusterNode", cluster: "Cluster", check_interval: time
         self._check_interval = check_interval
 
     async def _monitor(self) -> None:
+        from ray_on_golem.server.cluster.nodes import HeadClusterNode
+
         while True:
             activity_state = await self._node.activity.get_state()
 
@@ -139,6 +141,8 @@ def __init__(
         self._max_fail_count = max_fail_count
 
     async def _monitor(self) -> None:
+        from ray_on_golem.server.cluster.nodes import HeadClusterNode
+
         fails_count = 0
         while True:
             try: