gravel: remove etcd

Signed-off-by: Tim Serong <tserong@suse.com>

doc/dev/gravel/systemd-boot.md

@@ -17,7 +17,7 @@ These are
 
 * `/var/log`, so logs are persisted between runs.
 
-* `/var/lib/etcd`, where we keep this node's etcd persistent db.
+* `/var/lib/aquarium`, where we keep this node's local kvstore cache.
 
 * `/var/lib/containers`, where container images are kept.
 

doc/project-plan/roadmap.rst

@@ -73,7 +73,7 @@ M6
 
     * Add events widget (frontend)
 
-    * Basic event stashing on etcd (backend)
+    * Basic event stashing (backend)
       * node join
       * ???
 
@@ -171,7 +171,7 @@ A list of items brought to you in some unkempt, roughly prioritized order.
 
   * Obtain cluster events (backend)
 
-    * Store them in etcd?
+    * Store them somewhere?
     * Mutual exclusion access to ceph cluster operations
 
   * Dashboard
@@ -185,7 +185,7 @@ A list of items brought to you in some unkempt, roughly prioritized order.
 
       * Figure out what is an event (backend)
       * Figure out how to display Ceph status updates as events (backend)
-      * Store events in etcd (backend)
+      * Store events (backend)
       * Display events (frontend)
 
     * Hosts
@@ -199,7 +199,7 @@ A list of items brought to you in some unkempt, roughly prioritized order.
       * Obtain logs for each node (frontend, backend)
       * Obtain all logs (frontend, backend)
 
-      * Likely rely on etcd to do keep obtained logs from all nodes (backend)
+      * Where to keep obtained logs from all nodes? (backend)
 
         * or on websockets to connect to each node and obtain those logs?
 

doc/project-plan/testing-plan.rst

@@ -10,7 +10,7 @@ ceph tests should either be sufficient in this regard or expanded upon.
 Our primary scope is Aquarium. That is, we want to ensure aquarium does the
 right things, makes the right/expected decisions, and executes the orchestrated
 plan as expected. We also want to test the integration between Aquarium, the
-host OS, Ceph, and any other services (etcd for example) to ensure that the
+host OS, Ceph, and any other services to ensure that the
 deployment and operation of such services is functional and works with
 Aquarium's directives (ie, opinionated configuration).
 
@@ -130,4 +130,4 @@ aqrtest
 
 This is not run as part of any CI yet.
 
-TODO
+TODO

images/README.md

@@ -24,7 +24,7 @@ will be then be started on boot.
 The latter, `disk.sh`, is run within a chrooted image mount, but before the
 image is finalized. During this step we will be obtaining container images
 needed for Aquarium's execution, so we have them available upon first run,
-including an image for `Ceph` and an image for `etcd`.
+including an image for `Ceph`.
 
 
 ## Containerized build environment

images/aquarium/config.sh

@@ -81,8 +81,7 @@ if [[ "$kiwi_profiles" == *"Vagrant"* ]]; then
 fi
 
 pip install fastapi==0.63.0 uvicorn==0.13.3 websockets==8.1 \
-    bcrypt==3.2.0 pyjwt==2.1.0 python-multipart==0.0.5 \
-    git+https://github.com/aquarist-labs/aetcd3/@edf633045ce61c7bbac4d4a6ca15b14f8acfe9cd
+    bcrypt==3.2.0 pyjwt==2.1.0 python-multipart==0.0.5
 baseInsertService aquarium-boot
 baseInsertService sshd
 baseInsertService aquarium

images/aquarium/config.xml

@@ -179,7 +179,6 @@
         <package name="python38-requests"/>
         <package name="python38-PyYAML"/>
         <package name="python3-rados"/>
-        <package name="etcdctl"/>
         <archive name="aquarium.tar.gz"/>
         <archive name="root.tar.gz"/>
     </packages>

images/aquarium/disk.sh

@@ -39,7 +39,6 @@ mount -t tmpfs none /run
 
 # setting "--events-backend none" means podman doesn't try
 # (and fail) to log a "system refresh" event to the journal
-/usr/bin/podman --events-backend none pull quay.io/coreos/etcd:latest
 # we don't get to use cephadm directly because it will
 # try running a container inside the chroot, and that
 # fails with a bang.

src/boot/aqrbootsetup.sh

@@ -50,7 +50,6 @@ mount /dev/mapper/aquarium-systemdisk /aquarium
 # overlay
 overlay /etc etc || exit 1
 overlay /var/log logs || exit 1
-overlay /var/lib/etcd etcd || exit 1
 overlay /var/lib/aquarium aquarium || exit 1
 overlay /var/lib/containers containers || exit 1
 overlay /root roothome || exit 1

src/gravel/controllers/config.py

@@ -53,14 +53,6 @@ class ServicesOptionsModel(BaseModel):
     probe_interval: float = Field(1.0, title="Services Probe Interval")
 
 
-class EtcdOptionsModel(BaseModel):
-    registry: str = Field(
-        "quay.io/coreos/etcd", title="Container Image Registry"
-    )
-    version: str = Field("latest", title="Container Version Label")
-    data_dir: str = Field("/var/lib/etcd", title="Etcd Data Dir")
-
-
 class AuthOptionsModel(BaseModel):
     jwt_secret: str = Field(
         title="The access token secret",
@@ -81,7 +73,6 @@ class OptionsModel(BaseModel):
     devices: DevicesOptionsModel = Field(DevicesOptionsModel())
     status: StatusOptionsModel = Field(StatusOptionsModel())
     services: ServicesOptionsModel = Field(ServicesOptionsModel())
-    etcd: EtcdOptionsModel = Field(EtcdOptionsModel())
     auth: AuthOptionsModel = Field(AuthOptionsModel())
 
 

src/gravel/controllers/nodes/deployment.py

@@ -36,7 +36,6 @@
     NodeHasBeenDeployedError,
     NodeHasJoinedError,
 )
-from gravel.controllers.nodes.etcd import spawn_etcd
 from gravel.controllers.nodes.host import HostnameCtlError, set_hostname
 from gravel.controllers.nodes.messages import (
     ErrorMessageModel,
@@ -307,7 +306,6 @@ async def join(
         assert welcome.pubkey
         assert welcome.cephconf
         assert welcome.keyring
-        assert welcome.etcd_peer
 
         # create system disk after we are certain we are joining.
         # ensure all state writes happen only after the disk has been created.
@@ -321,17 +319,6 @@ async def join(
         self._state.mark_join()
         await self._set_hostname(hostname)
 
-        my_url: str = f"{hostname}=http://{address}:2380"
-        initial_cluster: str = f"{welcome.etcd_peer},{my_url}"
-        await spawn_etcd(
-            self._gstate,
-            new=False,
-            token=None,
-            hostname=hostname,
-            address=address,
-            initial_cluster=initial_cluster,
-        )
-
         authorized_keys: Path = Path("/root/.ssh/authorized_keys")
         if not authorized_keys.parent.exists():
             authorized_keys.parent.mkdir(0o700)
@@ -380,23 +367,6 @@ async def join(
         self._state.mark_ready()
         return True
 
-    async def _prepare_etcd(
-        self, hostname: str, address: str, token: str
-    ) -> None:
-        assert self._state
-        if self._state.bootstrapping:
-            raise NodeCantDeployError("node being deployed")
-        elif not self._state.nostage:
-            raise NodeCantDeployError("node can't be deployed")
-
-        await spawn_etcd(
-            self._gstate,
-            new=True,
-            token=token,
-            hostname=hostname,
-            address=address,
-        )
-
     async def deploy(
         self,
         config: DeploymentConfig,
@@ -485,11 +455,6 @@ async def _assimilate_devices() -> None:
         self._progress = ProgressEnum.PREPARING
 
         await self._set_hostname(hostname)
-        try:
-            await self._prepare_etcd(hostname, address, token)
-        except NodeError as e:
-            logger.error(f"bootstrap prepare error: {e.message}")
-            raise e
 
         await self._set_ntp_addr(ntp_addr)
 

src/gravel/controllers/nodes/messages.py

@@ -41,7 +41,6 @@ class WelcomeMessageModel(BaseModel):
     pubkey: str
     cephconf: str
     keyring: str
-    etcd_peer: str
 
 
 class ReadyToAddMessageModel(BaseModel):