golem-vm-provider 0.1.60__tar.gz → 0.1.61__tar.gz

{golem_vm_provider-0.1.60 → golem_vm_provider-0.1.61}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: golem-vm-provider
-Version: 0.1.60
+Version: 0.1.61
 Summary: VM on Golem Provider Node - Run your own provider node to offer VMs on the Golem Network
 Keywords: golem,vm,provider,cloud,decentralized
 Author: Phillip Jensen
@@ -456,7 +456,7 @@ Response (per stream):
 
 Notes:
 - Endpoints return 400 when streaming is disabled (zero `STREAM_PAYMENT_ADDRESS`).
-- In development mode (`GOLEM_PROVIDER_ENVIRONMENT=development`) additional debug logs are emitted around stream verification and monitor ticks.
+- In development mode (`GOLEM_ENVIRONMENT=development`) additional debug logs are emitted around stream verification and monitor ticks.
 
  ## Operations
  
@@ -467,7 +467,7 @@ Notes:
 golem-provider start
 
 # Development mode with extra logs and reload
-GOLEM_PROVIDER_ENVIRONMENT=development golem-provider start --network testnet
+GOLEM_ENVIRONMENT=development golem-provider start --network testnet
 ```
 
 Run as a background service (no terminal):
@@ -485,7 +485,7 @@ golem-provider status [--json]
 
 ### Mode vs. Network
 
-- Development Mode (`GOLEM_PROVIDER_ENVIRONMENT=development`)
+- Development Mode (`GOLEM_ENVIRONMENT=development`)
   - Optimizes for local iteration: enables reload + debug logging and uses local defaults (e.g., local port check servers). May derive a local/LAN IP automatically and prefix the provider name with `DEVMODE-`.
   - Does not decide which chain you target.
 
@@ -498,7 +498,7 @@ golem-provider status [--json]
   - Selects the payments chain profile (e.g., `l2.holesky`, `mainnet`). Determines default payments RPC, faucet enablement, and symbols.
 
 Common setups:
-- Local dev on testnet: `GOLEM_PROVIDER_ENVIRONMENT=development` plus `--network testnet`.
+- Local dev on testnet: `GOLEM_ENVIRONMENT=development` plus `--network testnet`.
 - Staging on testnet: keep `ENVIRONMENT=production`, set `--network testnet` and testnet RPCs.
 - Production on mainnet: `ENVIRONMENT=production` with `--network mainnet` and mainnet RPCs.
 

{golem_vm_provider-0.1.60 → golem_vm_provider-0.1.61}/README.md

@@ -411,7 +411,7 @@ Response (per stream):
 
 Notes:
 - Endpoints return 400 when streaming is disabled (zero `STREAM_PAYMENT_ADDRESS`).
-- In development mode (`GOLEM_PROVIDER_ENVIRONMENT=development`) additional debug logs are emitted around stream verification and monitor ticks.
+- In development mode (`GOLEM_ENVIRONMENT=development`) additional debug logs are emitted around stream verification and monitor ticks.
 
  ## Operations
  
@@ -422,7 +422,7 @@ Notes:
 golem-provider start
 
 # Development mode with extra logs and reload
-GOLEM_PROVIDER_ENVIRONMENT=development golem-provider start --network testnet
+GOLEM_ENVIRONMENT=development golem-provider start --network testnet
 ```
 
 Run as a background service (no terminal):
@@ -440,7 +440,7 @@ golem-provider status [--json]
 
 ### Mode vs. Network
 
-- Development Mode (`GOLEM_PROVIDER_ENVIRONMENT=development`)
+- Development Mode (`GOLEM_ENVIRONMENT=development`)
   - Optimizes for local iteration: enables reload + debug logging and uses local defaults (e.g., local port check servers). May derive a local/LAN IP automatically and prefix the provider name with `DEVMODE-`.
   - Does not decide which chain you target.
 
@@ -453,7 +453,7 @@ golem-provider status [--json]
   - Selects the payments chain profile (e.g., `l2.holesky`, `mainnet`). Determines default payments RPC, faucet enablement, and symbols.
 
 Common setups:
-- Local dev on testnet: `GOLEM_PROVIDER_ENVIRONMENT=development` plus `--network testnet`.
+- Local dev on testnet: `GOLEM_ENVIRONMENT=development` plus `--network testnet`.
 - Staging on testnet: keep `ENVIRONMENT=production`, set `--network testnet` and testnet RPCs.
 - Production on mainnet: `ENVIRONMENT=production` with `--network mainnet` and mainnet RPCs.
 

{golem_vm_provider-0.1.60 → golem_vm_provider-0.1.61}/provider/config.py

@@ -27,7 +27,7 @@ def ensure_config() -> None:
             created = True
 
     if not env_file.exists():
-        env_file.write_text("GOLEM_PROVIDER_ENVIRONMENT=production\n")
+        env_file.write_text("GOLEM_ENVIRONMENT=production\n")
         created = True
 
     from .security.ethereum import EthereumIdentity
@@ -55,8 +55,12 @@ class Settings(BaseSettings):
     PORT: int = 7466
     SKIP_PORT_VERIFICATION: bool = False
     ENVIRONMENT: str = "production"
-    # Logical network selector for annotation and client defaults
-    NETWORK: str = "mainnet"  # one of: "testnet", "mainnet"
+    # Logical network selector for advertisement scope and client defaults
+    # If not explicitly provided, computed by validator below (dev -> testnet, else -> mainnet)
+    NETWORK: str = Field(
+        default="",
+        description="Logical Golem network: 'testnet' or 'mainnet'"
+    )
 
     # Payments chain selection (modular network profiles). Keep default on l2.holesky
     PAYMENTS_NETWORK: str = Field(
@@ -73,6 +77,37 @@ class Settings(BaseSettings):
     def DEV_MODE(self) -> bool:
         return self.ENVIRONMENT == "development"
 
+    @field_validator("ENVIRONMENT", mode='before')
+    @classmethod
+    def prefer_global_env(cls, v: str) -> str:
+        """Prefer unified GOLEM_ENVIRONMENT when provided; fallback to service-specific env."""
+        ge = os.environ.get("GOLEM_ENVIRONMENT")
+        if ge:
+            return ge
+        return v
+
+    @field_validator("NETWORK", mode='before')
+    @classmethod
+    def resolve_network(cls, v: str, values: dict) -> str:
+        """Resolve logical network with sensible defaults.
+
+        Priority:
+        1) Explicit override via GOLEM_PROVIDER_NETWORK env or provided value
+        2) If ENVIRONMENT == development -> 'testnet'
+        3) Otherwise -> 'mainnet'
+        """
+        # Prefer explicit provider-scoped env override
+        env_override = os.environ.get("GOLEM_PROVIDER_NETWORK")
+        if env_override:
+            return env_override
+        # If value provided (via settings or direct assignment), keep it
+        val = (v or "").strip()
+        if val:
+            return val
+        # Default based on environment
+        env = (values.data.get("ENVIRONMENT") or "").lower()
+        return "testnet" if env == "development" else "mainnet"
+
     @field_validator("SKIP_PORT_VERIFICATION", mode='before')
     def set_skip_verification(cls, v: bool, values: dict) -> bool:
         """Set skip verification based on debug mode."""
@@ -186,6 +221,16 @@ class Settings(BaseSettings):
         description="Min withdrawable amount (wei) before triggering withdraw"
     )
 
+    # Behavior on exhausted runway
+    STREAM_REMOVE_MAPPING_ON_EXHAUSTED: bool = Field(
+        default=True,
+        description="When true, remove the VM->stream mapping after a successful stop on exhausted runway to prevent repeated stop attempts."
+    )
+    STREAM_DELETE_ON_EXHAUSTED: bool = Field(
+        default=False,
+        description="When true, delete the VM entirely once runway is exhausted and the VM has been stopped."
+    )
+
     # Shutdown behavior
     STOP_VMS_ON_EXIT: bool = Field(
         default=False,

{golem_vm_provider-0.1.60 → golem_vm_provider-0.1.61}/provider/discovery/golem_base_advertiser.py

@@ -1,7 +1,14 @@
 import asyncio
 from typing import Optional
 
-from golem_base_sdk import GolemBaseClient, GolemBaseCreate, GolemBaseUpdate, GolemBaseDelete, Annotation
+from golem_base_sdk import (
+    GolemBaseClient,
+    GolemBaseCreate,
+    GolemBaseUpdate,
+    GolemBaseDelete,
+    Annotation,
+    GolemBaseExtend,
+)
 from .advertiser import Advertiser
 from .golem_base_utils import get_provider_entity_keys
 from ..config import settings
@@ -36,7 +43,9 @@ class GolemBaseAdvertiser(Advertiser):
         try:
             while not self._stop_event.is_set():
                 await self.post_advertisement()
-                await asyncio.sleep(settings.GOLEM_BASE_ADVERTISEMENT_INTERVAL)
+                # Check more frequently than full TTL to ensure seamless renewal
+                interval = max(30, int(settings.GOLEM_BASE_ADVERTISEMENT_INTERVAL // 3))
+                await asyncio.sleep(interval)
         finally:
             await self.stop()
 
@@ -122,14 +131,36 @@ class GolemBaseAdvertiser(Advertiser):
                 logger.info(f"Current on-chain annotations: {current_annotations}")
                 logger.info(f"Expected annotations based on current config: {expected_annotations}")
 
+                desired_btl_blocks = int(settings.GOLEM_BASE_ADVERTISEMENT_INTERVAL) * 2  # ~2 blocks/sec
+
                 if sorted(current_annotations.items()) == sorted(expected_annotations.items()):
-                    logger.info("Advertisement is up-to-date. Waiting for expiration.")
+                    # Refresh TTL proactively if nearing expiry
+                    try:
+                        current_block = await self.client.http_client().eth.get_block_number()
+                        remaining_blocks = int(metadata.expires_at_block) - int(current_block)
+                    except Exception:
+                        # If we cannot determine remaining blocks, extend defensively
+                        remaining_blocks = 0
+
+                    # Refresh when below 20% of desired TTL (or if unknown/negative)
+                    refresh_threshold = max(10, desired_btl_blocks // 5)
+                    if remaining_blocks <= refresh_threshold:
+                        logger.info(
+                            f"Extending advertisement TTL (remaining_blocks={remaining_blocks}, threshold={refresh_threshold})."
+                        )
+                        ext = GolemBaseExtend(entity_key=entity_key, number_of_blocks=desired_btl_blocks)
+                        await self.client.extend_entities([ext])
+                        logger.info(f"Extended advertisement. Entity key: {entity_key}")
+                    else:
+                        logger.info(
+                            f"Advertisement up-to-date; TTL sufficient (remaining_blocks={remaining_blocks})."
+                        )
                 else:
                     logger.info("Advertisement is outdated. Updating.")
                     update = GolemBaseUpdate(
                         entity_key=entity_key,
                         data=b"",
-                        btl=settings.ADVERTISEMENT_INTERVAL * 2,
+                        btl=desired_btl_blocks,
                         string_annotations=string_annotations,
                         numeric_annotations=numeric_annotations,
                     )
@@ -144,9 +175,10 @@ class GolemBaseAdvertiser(Advertiser):
 
     async def _create_advertisement(self, string_annotations, numeric_annotations):
         """Helper to create a new advertisement."""
+        desired_btl_blocks = int(settings.GOLEM_BASE_ADVERTISEMENT_INTERVAL) * 2  # ~2 blocks/sec
         entity = GolemBaseCreate(
             data=b"",
-            btl=settings.ADVERTISEMENT_INTERVAL * 2,
+            btl=desired_btl_blocks,
             string_annotations=string_annotations,
             numeric_annotations=numeric_annotations,
         )

{golem_vm_provider-0.1.60 → golem_vm_provider-0.1.61}/provider/main.py

@@ -325,8 +325,8 @@ def status(json_out: bool = typer.Option(False, "--json", help="Output machine-r
     latest = _get_latest_version_from_pypi(pkg)
     update_available = bool(latest and current != latest)
 
-    # Environment
-    env = os.environ.get("GOLEM_PROVIDER_ENVIRONMENT", _settings.ENVIRONMENT)
+    # Environment (use unified GOLEM_ENVIRONMENT only)
+    env = os.environ.get("GOLEM_ENVIRONMENT", _settings.ENVIRONMENT)
     net = getattr(_settings, "NETWORK", None)
     dev_mode = env == "development" or bool(getattr(_settings, "DEV_MODE", False))
 
@@ -1324,7 +1324,7 @@ def stop(timeout: int = typer.Option(15, "--timeout", help="Seconds to wait for
     _remove_pid_file()
     print("Provider stopped")
 
-# Removed separate 'dev' command; use environment GOLEM_PROVIDER_ENVIRONMENT=development instead.
+# Removed separate 'dev' command; use environment GOLEM_ENVIRONMENT=development instead.
 
 def _env_path_for(dev_mode: Optional[bool]) -> str:
     from pathlib import Path
@@ -1535,9 +1535,10 @@ def run_server(
     from pathlib import Path
     from dotenv import load_dotenv
     import uvicorn
-    # Decide dev mode from explicit arg or environment
+    # Decide dev mode from explicit arg or environment (unified only)
     if dev_mode is None:
-        dev_mode = os.environ.get("GOLEM_PROVIDER_ENVIRONMENT", "").lower() == "development"
+        env_val = os.environ.get("GOLEM_ENVIRONMENT", "")
+        dev_mode = env_val.lower() == "development"
 
     # Load appropriate .env file based on mode
     env_file = ".env.dev" if dev_mode else ".env"

{golem_vm_provider-0.1.60 → golem_vm_provider-0.1.61}/provider/payments/monitor.py

@@ -54,7 +54,18 @@ class StreamMonitor:
                     try:
                         s = self.reader.get_stream(stream_id)
                     except Exception as e:
-                        logger.warning(f"stream {stream_id} lookup failed: {e}")
+                        # No payment info available; delete the VM and remove mapping per unified policy
+                        logger.info(
+                            f"Deleting VM {vm_id} due to missing/unavailable payment stream (id={stream_id}): {e}"
+                        )
+                        try:
+                            await self.vm_service.delete_vm(vm_id)
+                        except Exception as del_err:
+                            logger.warning(f"delete_vm failed for {vm_id} after stream lookup failure: {del_err}")
+                        try:
+                            await self.stream_map.remove(vm_id)
+                        except Exception as rem_err:
+                            logger.debug(f"failed to remove vm {vm_id} from stream map: {rem_err}")
                         continue
                     # Stop VM if remaining runway < threshold
                     remaining = max(int(s["stopTime"]) - int(now), 0)
@@ -69,33 +80,72 @@ class StreamMonitor:
                         )
                         try:
                             await self.vm_service.delete_vm(vm_id)
+                            # Best-effort verification of deletion for investigation
+                            try:
+                                _ = await self.vm_service.get_vm_status(vm_id)
+                                logger.info(
+                                    f"Post-delete status check: VM {vm_id} still present after delete request"
+                                )
+                            except VMNotFoundError:
+                                logger.info(
+                                    f"Post-delete status check: VM {vm_id} not found (expected)"
+                                )
+                            except Exception as chk_err:
+                                logger.debug(
+                                    f"Post-delete status check failed for {vm_id}: {chk_err}"
+                                )
                         except Exception as e:
                             logger.warning(f"delete_vm failed for {vm_id}: {e}")
                         try:
                             await self.stream_map.remove(vm_id)
+                            logger.debug(f"Removed {vm_id} from stream map after delete")
                         except Exception as e:
                             logger.debug(f"failed to remove vm {vm_id} from stream map: {e}")
                         continue
 
-                    # Only stop a VM when runway is completely empty
+                    # If runway is exhausted, delete the VM and remove mapping
                     if remaining == 0:
                         logger.info(
-                            f"Stopping VM {vm_id} as stream runway is exhausted (id={stream_id}, now={now}, stop={s.get('stopTime')})"
+                            f"Deleting VM {vm_id} as stream runway is exhausted (id={stream_id}, now={now}, stop={s.get('stopTime')})"
                         )
+                        # Capture pre-delete status for context
+                        try:
+                            pre = await self.vm_service.get_vm_status(vm_id)
+                            logger.info(
+                                f"Pre-delete status for {vm_id}: status={getattr(pre, 'status', '?')} ip={getattr(pre, 'ip_address', '?')}"
+                            )
+                        except VMNotFoundError:
+                            logger.info(
+                                f"Pre-delete status for {vm_id}: not found (will remove mapping)"
+                            )
+                        except Exception as pre_err:
+                            logger.debug(f"Pre-delete status check failed for {vm_id}: {pre_err}")
+
                         try:
-                            await self.vm_service.stop_vm(vm_id)
-                        except VMNotFoundError as e:
-                            # If the VM cannot be found, remove it from the stream map
-                            # to avoid repeated stop attempts and log spam.
-                            logger.warning(f"stop_vm failed for {vm_id}: {e}")
+                            await self.vm_service.delete_vm(vm_id)
+                            # Verify deletion
                             try:
-                                await self.stream_map.remove(vm_id)
-                            except Exception as rem_err:
+                                _ = await self.vm_service.get_vm_status(vm_id)
+                                logger.info(
+                                    f"Post-delete status check: VM {vm_id} still present after delete request"
+                                )
+                            except VMNotFoundError:
+                                logger.info(
+                                    f"Post-delete status check: VM {vm_id} not found (expected)"
+                                )
+                            except Exception as chk_err:
                                 logger.debug(
-                                    f"failed to remove vm {vm_id} from stream map after not-found: {rem_err}"
+                                    f"Post-delete status check failed for {vm_id}: {chk_err}"
                                 )
                         except Exception as e:
-                            logger.warning(f"stop_vm failed for {vm_id}: {e}")
+                            logger.warning(f"delete_vm failed for {vm_id}: {e}")
+                        try:
+                            await self.stream_map.remove(vm_id)
+                            logger.info(f"Removed mapping for {vm_id} after delete on exhausted runway")
+                        except Exception as rem_err:
+                            logger.debug(
+                                f"failed to remove vm {vm_id} from stream map after delete: {rem_err}"
+                            )
                         continue
 
                     # Otherwise, do not stop; just log health and consider withdrawals

{golem_vm_provider-0.1.60 → golem_vm_provider-0.1.61}/provider/vm/multipass_adapter.py

@@ -87,7 +87,8 @@ class MultipassAdapter(VMProvider):
         """Get detailed information about a VM."""
         try:
             result = await self._run_multipass(["info", vm_id, "--format", "json"])
-            logger.info(f"Raw multipass info for {vm_id}: {result.stdout}")
+            # Only log raw multipass output in debug mode to avoid noisy logs
+            logger.debug(f"Raw multipass info for {vm_id}: {result.stdout}")
             info = json.loads(result.stdout)
             vm_info = info["info"][vm_id]
             essential_fields = ["state", "ipv4", "cpu_count", "memory", "disks"]

{golem_vm_provider-0.1.60 → golem_vm_provider-0.1.61}/provider/vm/service.py

@@ -59,6 +59,7 @@ class VMService:
 
         try:
             vm_info = await self.provider.get_vm_status(multipass_name)
+            logger.info(f"Deleting VM {vm_id} (multipass={multipass_name}) with status={vm_info.status}")
             await self.provider.delete_vm(multipass_name)
             await self.resource_tracker.deallocate(vm_info.resources, vm_id)
             # Optional: best-effort on-chain termination if we have a mapping
@@ -83,7 +84,9 @@ class VMService:
         multipass_name = await self.name_mapper.get_multipass_name(vm_id)
         if not multipass_name:
             raise VMNotFoundError(f"VM {vm_id} not found")
+        logger.info(f"Stopping VM {vm_id} (multipass={multipass_name})")
         vm = await self.provider.stop_vm(multipass_name)
+        logger.info(f"Stopped VM {vm_id} result status={getattr(vm, 'status', '?')}")
         # Optional: best-effort withdraw for active stream
         try:
             if self.blockchain_client:

{golem_vm_provider-0.1.60 → golem_vm_provider-0.1.61}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "golem-vm-provider"
-version = "0.1.60"
+version = "0.1.61"
 description = "VM on Golem Provider Node - Run your own provider node to offer VMs on the Golem Network"
 authors = ["Phillip Jensen <phillip+vm-on-golem@golemgrid.com>"]
 readme = "README.md"