Merge pull request 'Add resource limit options to spec.' (#735) from telackey/limits into main

Reviewed-on: cerc-io/stack-orchestrator#735
Reviewed-by: David Boreham <dboreham@noreply.git.vdb.to>

requirements.txt

@@ -10,3 +10,4 @@ pydantic==1.10.9
 tomli==2.0.1
 validators==0.22.0
 kubernetes>=28.1.0
+humanfriendly>=10.0

stack_orchestrator/deploy/k8s/cluster_info.py

@@ -25,9 +25,38 @@ from stack_orchestrator.deploy.k8s.helpers import get_node_pv_mount_path
 from stack_orchestrator.deploy.k8s.helpers import envs_from_environment_variables_map
 from stack_orchestrator.deploy.deploy_util import parsed_pod_files_map_from_file_names, images_for_deployment
 from stack_orchestrator.deploy.deploy_types import DeployEnvVars
-from stack_orchestrator.deploy.spec import Spec
+from stack_orchestrator.deploy.spec import Spec, Resources, ResourceLimits
 from stack_orchestrator.deploy.images import remote_tag_for_image
 
+DEFAULT_VOLUME_RESOURCES = Resources({
+    "reservations": {"storage": "2Gi"}
+})
+
+DEFAULT_CONTAINER_RESOURCES = Resources({
+    "reservations": {"cpus": "0.1", "memory": "200M"},
+    "limits": {"cpus": "1.0", "memory": "2000M"},
+})
+
+
+def to_k8s_resource_requirements(resources: Resources) -> client.V1ResourceRequirements:
+    def to_dict(limits: ResourceLimits):
+        if not limits:
+            return None
+
+        ret = {}
+        if limits.cpus:
+            ret["cpu"] = str(limits.cpus)
+        if limits.memory:
+            ret["memory"] = f"{int(limits.memory / (1000 * 1000))}M"
+        if limits.storage:
+            ret["storage"] = f"{int(limits.storage / (1000 * 1000))}M"
+        return ret
+
+    return client.V1ResourceRequirements(
+        requests=to_dict(resources.reservations),
+        limits=to_dict(resources.limits)
+    )
+
 
 class ClusterInfo:
     parsed_pod_yaml_map: Any
@@ -135,9 +164,13 @@ class ClusterInfo:
         result = []
         spec_volumes = self.spec.get_volumes()
         named_volumes = named_volumes_from_pod_files(self.parsed_pod_yaml_map)
+        resources = self.spec.get_volume_resources()
+        if not resources:
+            resources = DEFAULT_VOLUME_RESOURCES
         if opts.o.debug:
             print(f"Spec Volumes: {spec_volumes}")
             print(f"Named Volumes: {named_volumes}")
+            print(f"Resources: {resources}")
         for volume_name in spec_volumes:
             if volume_name not in named_volumes:
                 if opts.o.debug:
@@ -146,9 +179,7 @@ class ClusterInfo:
             spec = client.V1PersistentVolumeClaimSpec(
                 access_modes=["ReadWriteOnce"],
                 storage_class_name="manual",
-                resources=client.V1ResourceRequirements(
-                    requests={"storage": "2Gi"}
-                ),
+                resources=to_k8s_resource_requirements(resources),
                 volume_name=f"{self.app_name}-{volume_name}"
             )
             pvc = client.V1PersistentVolumeClaim(
@@ -192,6 +223,9 @@ class ClusterInfo:
         result = []
         spec_volumes = self.spec.get_volumes()
         named_volumes = named_volumes_from_pod_files(self.parsed_pod_yaml_map)
+        resources = self.spec.get_volume_resources()
+        if not resources:
+            resources = DEFAULT_VOLUME_RESOURCES
         for volume_name in spec_volumes:
             if volume_name not in named_volumes:
                 if opts.o.debug:
@@ -200,7 +234,7 @@ class ClusterInfo:
             spec = client.V1PersistentVolumeSpec(
                 storage_class_name="manual",
                 access_modes=["ReadWriteOnce"],
-                capacity={"storage": "2Gi"},
+                capacity=to_k8s_resource_requirements(resources).requests,
                 host_path=client.V1HostPathVolumeSource(path=get_node_pv_mount_path(volume_name))
             )
             pv = client.V1PersistentVolume(
@@ -214,6 +248,9 @@ class ClusterInfo:
     # TODO: put things like image pull policy into an object-scope struct
     def get_deployment(self, image_pull_policy: str = None):
         containers = []
+        resources = self.spec.get_container_resources()
+        if not resources:
+            resources = DEFAULT_CONTAINER_RESOURCES
         for pod_name in self.parsed_pod_yaml_map:
             pod = self.parsed_pod_yaml_map[pod_name]
             services = pod["services"]
@@ -237,10 +274,7 @@ class ClusterInfo:
                     env=envs_from_environment_variables_map(self.environment_variables.map),
                     ports=[client.V1ContainerPort(container_port=port)],
                     volume_mounts=volume_mounts,
-                    resources=client.V1ResourceRequirements(
-                        requests={"cpu": "100m", "memory": "200Mi"},
-                        limits={"cpu": "1000m", "memory": "2000Mi"},
-                    ),
+                    resources=to_k8s_resource_requirements(resources),
                 )
                 containers.append(container)
         volumes = volumes_for_pod_files(self.parsed_pod_yaml_map, self.spec, self.app_name)

stack_orchestrator/deploy/spec.py

@@ -13,12 +13,60 @@
 # You should have received a copy of the GNU Affero General Public License
 # along with this program.  If not, see <http:#www.gnu.org/licenses/>.
 
-from pathlib import Path
 import typing
+import humanfriendly
+
+from pathlib import Path
+
 from stack_orchestrator.util import get_yaml
 from stack_orchestrator import constants
 
 
+class ResourceLimits:
+    cpus: float = None
+    memory: int = None
+    storage: int = None
+
+    def __init__(self, obj={}):
+        if "cpus" in obj:
+            self.cpus = float(obj["cpus"])
+        if "memory" in obj:
+            self.memory = humanfriendly.parse_size(obj["memory"])
+        if "storage" in obj:
+            self.storage = humanfriendly.parse_size(obj["storage"])
+
+    def __len__(self):
+        return len(self.__dict__)
+
+    def __iter__(self):
+        for k in self.__dict__:
+            yield k, self.__dict__[k]
+
+    def __repr__(self):
+        return str(self.__dict__)
+
+
+class Resources:
+    limits: ResourceLimits = None
+    reservations: ResourceLimits = None
+
+    def __init__(self, obj={}):
+        if "reservations" in obj:
+            self.reservations = ResourceLimits(obj["reservations"])
+        if "limits" in obj:
+            self.limits = ResourceLimits(obj["limits"])
+
+    def __len__(self):
+        return len(self.__dict__)
+
+    def __iter__(self):
+        for k in self.__dict__:
+            yield k, self.__dict__[k]
+
+    def __repr__(self):
+        return str(self.__dict__)
+
+
 class Spec:
 
     obj: typing.Any
@@ -47,6 +95,12 @@ class Spec:
                 if self.obj and "configmaps" in self.obj
                 else {})
 
+    def get_container_resources(self):
+        return Resources(self.obj.get("resources", {}).get("containers", {}))
+
+    def get_volume_resources(self):
+        return Resources(self.obj.get("resources", {}).get("volumes", {}))
+
     def get_http_proxy(self):
         return (self.obj[constants.network_key][constants.http_proxy_key]
                 if self.obj and constants.network_key in self.obj

stack_orchestrator/deploy/webapp/deploy_webapp_from_registry.py

@@ -29,7 +29,7 @@ from stack_orchestrator.deploy.webapp.util import (LaconicRegistryClient,
                                                    build_container_image, push_container_image,
                                                    file_hash, deploy_to_k8s, publish_deployment,
                                                    hostname_for_deployment_request, generate_hostname_for_app,
-                                                   match_owner)
+                                                   match_owner, skip_by_tag)
 
 
 def process_app_deployment_request(
@@ -222,17 +222,6 @@ def command(ctx, kube_config, laconic_config, image_registry, deployment_parent_
             dump_known_requests(state_file, requests)
         return
 
-    def skip_by_tag(r):
-        for tag in exclude_tags:
-            if tag and r.attributes.tags and tag in r.attributes.tags:
-                return True
-
-        for tag in include_tags:
-            if tag and (not r.attributes.tags or tag not in r.attributes.tags):
-                return True
-
-        return False
-
     previous_requests = load_known_requests(state_file)
 
     # Collapse related requests.
@@ -256,7 +245,7 @@ def command(ctx, kube_config, laconic_config, image_registry, deployment_parent_
             print("Ignoring request %s, it has been superseded." % r.id)
             continue
 
-        if skip_by_tag(r):
+        if skip_by_tag(r, include_tags, exclude_tags):
             print("Skipping request %s, filtered by tag (include %s, exclude %s, present %s)" % (r.id,
                                                                                                  include_tags,
                                                                                                  exclude_tags,

stack_orchestrator/deploy/webapp/undeploy_webapp_from_registry.py

@@ -20,7 +20,7 @@ import sys
 
 import click
 
-from stack_orchestrator.deploy.webapp.util import LaconicRegistryClient, match_owner
+from stack_orchestrator.deploy.webapp.util import LaconicRegistryClient, match_owner, skip_by_tag
 
 
 def process_app_removal_request(ctx,
@@ -107,10 +107,12 @@ def dump_known_requests(filename, requests):
 @click.option("--delete-names/--preserve-names", help="Delete all names associated with removed deployments.", default=True)
 @click.option("--delete-volumes/--preserve-volumes", default=True, help="delete data volumes")
 @click.option("--dry-run", help="Don't do anything, just report what would be done.", is_flag=True)
+@click.option("--include-tags", help="Only include requests with matching tags (comma-separated).", default="")
+@click.option("--exclude-tags", help="Exclude requests with matching tags (comma-separated).", default="")
 @click.pass_context
 def command(ctx, laconic_config, deployment_parent_dir,
             request_id, discover, state_file, only_update_state,
-            delete_names, delete_volumes, dry_run):
+            delete_names, delete_volumes, dry_run, include_tags, exclude_tags):
     if request_id and discover:
         print("Cannot specify both --request-id and --discover", file=sys.stderr)
         sys.exit(2)
@@ -123,6 +125,10 @@ def command(ctx, laconic_config, deployment_parent_dir,
         print("--only-update-state requires --state-file", file=sys.stderr)
         sys.exit(2)
 
+    # Split CSV and clean up values.
+    include_tags = [tag.strip() for tag in include_tags.split(",") if tag]
+    exclude_tags = [tag.strip() for tag in exclude_tags.split(",") if tag]
+
     laconic = LaconicRegistryClient(laconic_config)
 
     # Find deployment removal requests.
@@ -155,10 +161,22 @@ def command(ctx, laconic_config, deployment_parent_dir,
             # TODO: should we handle CRNs?
             removals_by_deployment[r.attributes.deployment] = r
 
-    requests_to_execute = []
+    one_per_deployment = {}
     for r in requests:
         if not r.attributes.deployment:
             print(f"Skipping removal request {r.id} since it was a cancellation.")
+        elif r.attributes.deployment in one_per_deployment:
+            print(f"Skipping removal request {r.id} since it was superseded.")
+        else:
+            one_per_deployment[r.attributes.deployment] = r
+
+    requests_to_execute = []
+    for r in one_per_deployment.values():
+        if skip_by_tag(r, include_tags, exclude_tags):
+            print("Skipping removal request %s, filtered by tag (include %s, exclude %s, present %s)" % (r.id,
+                                                                                                         include_tags,
+                                                                                                         exclude_tags,
+                                                                                                         r.attributes.tags))
         elif r.id in removals_by_request:
             print(f"Found satisfied request for {r.id} at {removals_by_request[r.id].id}")
         elif r.attributes.deployment in removals_by_deployment:

stack_orchestrator/deploy/webapp/util.py

@@ -351,3 +351,15 @@ def generate_hostname_for_app(app):
     else:
         m.update(app.attributes.repository.encode())
     return "%s-%s" % (last_part, m.hexdigest()[0:10])
+
+
+def skip_by_tag(r, include_tags, exclude_tags):
+    for tag in exclude_tags:
+        if tag and r.attributes.tags and tag in r.attributes.tags:
+            return True
+
+    for tag in include_tags:
+        if tag and (not r.attributes.tags or tag not in r.attributes.tags):
+            return True
+
+    return False