List machine improvements (#365)

- Use free GPUs, not total GPUs when displaying GPUs available
- Sort GPUs in table/json view for consistency (backend does not have
ordered maps)
- Add shebang to agent bootstrap script for better copy/paste into
userdata of provider
- Prevent python-rich from cropping/"correcting" lines in agent
bootstrap script for reliable copy/paste

Resolve BE-1542

pkg/repository/provider_redis.go

@@ -340,19 +340,6 @@ func (r *ProviderRedisRepository) RemoveMachineLock(providerName, poolName, mach
 	return r.lock.Release(common.RedisKeys.ProviderMachineLock(providerName, poolName, machineId))
 }
 
-// GetGPUAvailability retrieves the GPU availability status for all machines
-// managed by various providers and pools.
-//
-// Usage:
-//
-//	availability, err := machineRepo.GetGPUAvailability(pools)
-//	if err != nil {
-//		log.Fatal(err)
-//	}
-//
-//	for gpu, isAvailable := range availability {
-//		fmt.Printf("GPU: %s, Available: %t\n", gpu, isAvailable)
-//	}
 func (r *ProviderRedisRepository) GetGPUAvailability(pools map[string]types.WorkerPoolConfig) (map[string]bool, error) {
 	gpuAvailability := map[string]bool{}
 
@@ -376,7 +363,9 @@ func (r *ProviderRedisRepository) GetGPUAvailability(pools map[string]types.Work
 
 		// Update availability of GPU based on machine states
 		for _, machine := range machines {
-			gpuAvailability[machine.State.Gpu] = machine.State.GpuCount > 0
+			if machine.Metrics != nil {
+				gpuAvailability[machine.State.Gpu] = machine.Metrics.FreeGpuCount > 0
+			}
 		}
 	}
 

sdk/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "beta9"
-version = "0.1.50"
+version = "0.1.51"
 description = ""
 authors = ["beam.cloud <support@beam.cloud>"]
 packages = [

sdk/src/beta9/cli/machine.py

@@ -1,3 +1,4 @@
+import textwrap
 from datetime import datetime, timezone
 
 import click
@@ -78,6 +79,8 @@ def list_machines(
     if not res.ok:
         terminal.error(res.err_msg)
 
+    res.gpus = {gpu: res.gpus[gpu] for gpu in sorted(res.gpus)}
+
     if format == "json":
         machines = [d.to_dict(casing=Casing.SNAKE) for d in res.machines]  # type:ignore
         terminal.print_json({"machines": machines, "gpus": res.gpus})
@@ -165,13 +168,19 @@ def create_machine(service: ServiceClient, pool: str):
             f"Created machine with ID: '{res.machine.id}'. Use the following command to setup the node:"
         )
         terminal.detail(
-            f"""sudo curl -L -o agent https://release.beam.cloud/agent/agent && \\
-sudo chmod +x agent && \\
-sudo ./agent --token "{res.machine.registration_token}" --machine-id "{res.machine.id}" \\
---tailscale-url "{res.machine.tailscale_url}" \\
---tailscale-auth "{res.machine.tailscale_auth}" \\
---pool-name "{res.machine.pool_name}" \\
---provider-name "{res.machine.provider_name}" """
+            textwrap.dedent(f"""
+            #!/bin/bash
+            sudo curl -L -o agent https://release.beam.cloud/agent/agent && \\
+            sudo chmod +x agent && \\
+            sudo ./agent --token "{res.machine.registration_token}" \\
+              --machine-id "{res.machine.id}" \\
+              --tailscale-url "{res.machine.tailscale_url}" \\
+              --tailscale-auth "{res.machine.tailscale_auth}" \\
+              --pool-name "{res.machine.pool_name}" \\
+              --provider-name "{res.machine.provider_name}"
+            """),
+            crop=False,
+            overflow="ignore",
         )
 
     else: