Configurable stacks

.github/workflows/ci.yml

@@ -79,7 +79,7 @@ jobs:
       - name: Create tarball of log archive
         if: success() || failure()
         run: |
-          tar -czf /tmp/artifacts/archive.tar ~/.local/share/ceph-devstack/archive/
+          tar -czf /tmp/artifacts/archive.tar ~/.local/share/ceph-devstack/teuthology/archive/
       - name: Upload log archive
         if: success() || failure()
         uses: actions/upload-artifact@v7.0.1

README.md

@@ -105,7 +105,8 @@ pool state file and its lock always stay on the local machine.
 Each container picks host loop devices at create time: it needs
 `loop_device_count` devices of `loop_device_size`, reuses any existing backing
 files for that container name, then takes the lowest free loop numbers after
-scanning what is already attached or claimed in the `disk_images` directory.
+scanning what is already attached or claimed in any stack's `disk_images`
+directory.
 
 If `block_pool.json` is lost but the on-disk marker remains, start again to
 reclaim the marker; the device is not reformatted. The old allocations are
@@ -146,7 +147,7 @@ Once everything is started, a message similar to this will be logged:
 
 `View test results at http://smithi065.front.sepia.ceph.com:8081/`
 
-This link points to the running Pulpito instance. Test archives are also stored in the `--data-dir` (default: `~/.local/share/ceph-devstack`).
+This link points to the running Pulpito instance. Test archives are also stored in `<data_dir>/teuthology/archive` (`data_dir` defaults to `~/.local/share/ceph-devstack`).
 
 To watch teuthology's output, you can:
 

ceph_devstack/__init__.py

@@ -42,6 +42,11 @@ def parse_args(args: List[str]) -> argparse.Namespace:
         default=DEFAULT_CONFIG_PATH,
         help="Path to the ceph-devstack config file",
     )
+    parser.add_argument(
+        "--stack",
+        default=None,
+        help="Stack to deploy (default: value of 'stack' in config)",
+    )
     subparsers = parser.add_subparsers(dest="command")
     parser_config = subparsers.add_parser("config", help="Get or set config items")
     subparsers_config = parser_config.add_subparsers(dest="config_op")
@@ -216,3 +221,8 @@ def unset_value(self, name: str) -> None:
 
 config = Config()
 config.load()
+
+
+def base_data_dir() -> Path:
+    """The configured data_dir; each stack keeps its state in a subdirectory."""
+    return Path(config.get("data_dir", DEFAULT_DATA_DIR)).expanduser().absolute()

ceph_devstack/cli.py

@@ -2,7 +2,8 @@
 import logging
 import sys
 
-from pathlib import Path
+from argparse import Namespace
+from subprocess import CalledProcessError
 
 from ceph_devstack import config, logger, parse_args, VERBOSE
 from ceph_devstack.block_pool import BlockPoolError, format_size, shared_pool
@@ -16,22 +17,6 @@
     "unset": lambda config, args: config.unset_value(args.name),
 }
 
-COMMAND_HANDLERS = {
-    "doctor": None,
-    "apply": lambda args, obj: obj.apply(args.command),
-    "pull": lambda _, obj: obj.pull(),
-    "build": lambda _, obj: obj.build(),
-    "create": lambda _, obj: obj.create(),
-    "remove": lambda _, obj: obj.remove(),
-    "start": lambda _, obj: obj.start(),
-    "stop": lambda _, obj: obj.stop(),
-    "watch": lambda _, obj: obj.watch(),
-    "wait": lambda args, obj: obj.wait(container_name=args.container),
-    "logs": lambda args, obj: obj.logs(
-        run_name=args.run_name, job_id=args.job_id, locate=args.locate
-    ),
-}
-
 
 def _configure_console_logging(verbose: bool) -> None:
     if not verbose:
@@ -67,6 +52,36 @@ def block_pool_status() -> int:
     return 0
 
 
+def _action_kwargs(args: Namespace) -> dict:
+    match args.command:
+        case "wait":
+            return {"container_name": args.container}
+        case "logs":
+            return {
+                "run_name": args.run_name,
+                "job_id": args.job_id,
+                "locate": args.locate,
+            }
+        case _:
+            return {}
+
+
+async def _apply(stack: CephDevStack, action: str, **kwargs) -> int:
+    """Check requirements, then run the action (if any) on the stack."""
+    if not all(
+        await asyncio.gather(
+            check_requirements(),
+            stack.check_requirements(),
+        )
+    ):
+        logger.error("Requirements not met!")
+        return 1
+    if action == "doctor":
+        return 0
+    result = await stack.apply(action, **kwargs)
+    return result if result is not None else 0
+
+
 def main() -> int:
     args = parse_args(sys.argv[1:])
     config.load(args.config_file)
@@ -80,25 +95,17 @@ def main() -> int:
             return 2
         return block_pool_status()
     config["args"] = vars(args)
-    data_path = Path(config["data_dir"]).expanduser()
-    data_path.mkdir(parents=True, exist_ok=True)
-    obj = CephDevStack()
-
-    async def run():
-        if not all(
-            await asyncio.gather(
-                check_requirements(),
-                obj.check_requirements(),
-            )
-        ):
-            logger.error("Requirements not met!")
-            return 1
-        handler = COMMAND_HANDLERS.get(args.command)
-        if handler:
-            return await handler(args, obj)
-
     try:
-        sys.exit(asyncio.run(run()))
+        stack = CephDevStack(stack_name=args.stack)
+    except ValueError as exc:
+        logger.error(str(exc))
+        return 1
+    stack.data_dir.mkdir(parents=True, exist_ok=True)
+    action = args.command
+    try:
+        return asyncio.run(_apply(stack, action, **_action_kwargs(args)))
+    except CalledProcessError as exc:
+        return exc.returncode if exc.returncode is not None else 1
     except KeyboardInterrupt:
         logger.debug("Exiting!")
         return 130  # 128 + SIGINT

ceph_devstack/config.toml

@@ -1,3 +1,4 @@
+stack = "teuthology"
 data_dir = "~/.local/share/ceph-devstack"
 
 [block_pool]
@@ -14,6 +15,18 @@ data_dir = "~/.local/share/ceph-devstack"
 # allow_enroll = true
 # state_dir = "/path"  # where block_pool.json lives; defaults to data_dir
 
+[stacks.teuthology]
+services = [
+    "postgres",
+    "paddles",
+    "beanstalk",
+    "pulpito",
+    "teuthology",
+    "testnode",
+    "archive",
+]
+secrets = ["ssh_keypair"]
+
 [containers.archive]
 image = "python:alpine"
 

ceph_devstack/resources/__init__.py

@@ -6,12 +6,42 @@
 
 from pathlib import Path
 from subprocess import CalledProcessError
-from typing import List, Dict, Set
+from typing import List, Dict, Optional, Protocol, Set
 
+from ceph_devstack import config
 from ceph_devstack.exec import Subprocess
 from ceph_devstack.host import host, local_host
 
 
+class StackResource(Protocol):
+    """Interface that CephDevStack expects from every service it manages."""
+
+    def __init__(
+        self, name: str = "", *, data_dir: Path, containers_config: dict
+    ) -> None: ...
+
+    @property
+    def name(self) -> str: ...
+
+    async def pull(self) -> None: ...
+
+    async def build(self) -> None: ...
+
+    async def create(self) -> None: ...
+
+    async def start(self) -> None: ...
+
+    async def stop(self) -> None: ...
+
+    async def remove(self) -> None: ...
+
+    async def exists(self) -> bool: ...
+
+    async def is_running(self) -> bool: ...
+
+    async def wait(self) -> int | None: ...
+
+
 class DevStack:
     def __init__(self, args: argparse.Namespace):
         self.args = args
@@ -40,16 +70,38 @@ class PodmanResource:
     log: Dict[str, Set[str]] = {}
     _name: str | None = None
 
-    def __init__(self, name: str = ""):
+    def __init__(
+        self,
+        name: str = "",
+        *,
+        data_dir: Optional[Path] = None,
+        containers_config: Optional[dict] = None,
+    ):
         if name:
             self._name = name
+        self._data_dir = data_dir
+        self._containers_config = containers_config
 
     @property
     def name(self) -> str:
         if self._name is not None:
             return self._name
         return self.__class__.__name__.lower()
 
+    @property
+    def data_dir(self) -> Path:
+        """Where this resource keeps its on-disk state; set by its stack."""
+        if self._data_dir is None:
+            raise ValueError(f"{self.name}: no data_dir was given")
+        return self._data_dir
+
+    @property
+    def containers_config(self) -> dict:
+        """The stack's merged [containers] config, else the global one."""
+        if self._containers_config is None:
+            return config["containers"]
+        return self._containers_config
+
     async def cmd(
         self,
         args: List[str],