deployer: add info generating commands under resource-allocation

deployer/README.md

@@ -264,6 +264,12 @@ This sub-command can be used to generate the resource allocation choices for giv
 ##### `generate resource-allocation choices`
 This generates a custom number of resource allocation choices for a certain instance type, depending on a certain chosen strategy that can be used in the profile list of a hub.
 
+##### `generate resource-allocation daemonset-overhead`
+Updates `daemonset_overhead.yaml` with an individual cluster's DaemonSets with running pods combined requests of CPU and memory, excluding GPU related DaemonSets.
+
+##### `generate resource-allocation instance-capacity`
+Updates `instance_capacity.yaml` with instance types' total and allocatable capacity reported by `kubectl get node`.
+
 ##### `generate resource-allocation node-info-update`
 This updates the json file `node-capacity-info.json` with info about the capacity of a node of a certain type. This file is then used for generating the resource choices.
 

deployer/__main__.py

@@ -9,7 +9,9 @@
 import deployer.commands.generate.dedicated_cluster.aws  # noqa: F401
 import deployer.commands.generate.dedicated_cluster.gcp  # noqa: F401
 import deployer.commands.generate.helm_upgrade.jobs  # noqa: F401
+import deployer.commands.generate.resource_allocation.daemonset_overhead  # noqa: F401
 import deployer.commands.generate.resource_allocation.generate_choices  # noqa: F401
+import deployer.commands.generate.resource_allocation.instance_capacity  # noqa: F401
 import deployer.commands.generate.resource_allocation.update_nodeinfo  # noqa: F401
 import deployer.commands.grafana.central_grafana  # noqa: F401
 import deployer.commands.grafana.deploy_dashboards  # noqa: F401

deployer/commands/generate/resource_allocation/daemonset_overhead.py

@@ -0,0 +1,152 @@
+import json
+import math
+import subprocess
+from pathlib import Path
+
+import typer
+from kubernetes.utils.quantity import parse_quantity
+from ruamel.yaml import YAML
+
+from deployer.infra_components.cluster import Cluster
+from deployer.utils.file_acquisition import find_absolute_path_to_cluster_file
+
+from .resource_allocation_app import resource_allocation_app
+
+HERE = Path(__file__).parent
+yaml = YAML()
+yaml.preserve_quotes = True
+yaml.indent(mapping=2, sequence=4, offset=2)
+
+
+def get_k8s_distribution():
+    """
+    Returns a 2-tuple with the guessed the k8s distribution based on the k8s
+    api-server's reported version, either Google's GKE, Amazon's EKS, or Azure's
+    AKS, and the server's reported gitVersion.
+    """
+    output = subprocess.check_output(
+        [
+            "kubectl",
+            "version",
+            "--output=json",
+        ],
+        text=True,
+    )
+    version_info = json.loads(output)
+    server_version_info = version_info["serverVersion"]["gitVersion"]
+    if "gke" in server_version_info:
+        return "gke", server_version_info
+    if "eks" in server_version_info:
+        return "eks", server_version_info
+    else:
+        return "aks", server_version_info
+
+
+def get_daemon_sets_requests():
+    """
+    Returns a list of info about DaemonSets with pods desired to be scheduled on
+    some nodes the k8s cluster.
+    """
+    output = subprocess.check_output(
+        [
+            "kubectl",
+            "get",
+            "ds",
+            "--all-namespaces",
+            "--output=jsonpath-as-json={.items[*]}",
+        ],
+        text=True,
+    )
+    daemon_sets = json.loads(output)
+
+    # filter out DaemonSets that aren't desired on any node
+    daemon_sets = [ds for ds in daemon_sets if ds["status"]["desiredNumberScheduled"]]
+
+    info = []
+    for ds in daemon_sets:
+        name = ds["metadata"]["name"]
+        req_mem = req_cpu = lim_mem = lim_cpu = 0
+        for c in ds["spec"]["template"]["spec"]["containers"]:
+            resources = c.get("resources", {})
+            requests = resources.get("requests", {})
+            limits = resources.get("limits", {})
+            req_mem += parse_quantity(requests.get("memory", 0))
+            lim_mem += parse_quantity(limits.get("memory", 0))
+            req_cpu += parse_quantity(requests.get("cpu", 0))
+            lim_cpu += parse_quantity(limits.get("cpu", 0))
+
+        info.append(
+            {
+                "name": name,
+                "cpu_request": float(req_cpu),
+                "cpu_limit": float(lim_cpu),
+                "memory_request": int(req_mem),
+                "memory_limit": int(lim_mem),
+            }
+        )
+
+    return info
+
+
+def get_daemon_sets_overhead():
+    """
+    Returns a summary of the overhead from get_daemon_sets_requests.
+    """
+    daemon_sets = get_daemon_sets_requests()
+    # filter out DaemonSets related to nvidia GPUs
+    daemon_sets = [ds for ds in daemon_sets if "nvidia" not in ds["name"]]
+    # separate DaemonSets without requests, as only requests are what impacts
+    # scheduling of pods and reduces a node's remaining allocatable resources
+    req_daemon_sets = [
+        ds for ds in daemon_sets if ds["cpu_request"] or ds["memory_request"]
+    ]
+    other_daemon_sets = [
+        ds for ds in daemon_sets if not ds["cpu_request"] and not ds["memory_request"]
+    ]
+
+    cpu_requests = sum([ds["cpu_request"] for ds in req_daemon_sets])
+    memory_requests = sum([ds["memory_request"] for ds in req_daemon_sets])
+    info = {
+        "requesting_daemon_sets": ",".join(
+            sorted([ds["name"] for ds in req_daemon_sets])
+        ),
+        "other_daemon_sets": ",".join(sorted([ds["name"] for ds in other_daemon_sets])),
+        "cpu_requests": str(math.ceil(cpu_requests * 1000)) + "m",
+        "memory_requests": str(math.ceil(memory_requests / 1024**2)) + "Mi",
+    }
+    return info
+
+
+@resource_allocation_app.command()
+def daemonset_overhead(
+    cluster_name: str = typer.Argument(..., help="Name of cluster to operate on"),
+):
+    """
+    Updates `daemonset_overhead.yaml` with an individual cluster's DaemonSets
+    with running pods combined requests of CPU and memory, excluding GPU related
+    DaemonSets.
+    """
+    file_path = HERE / "daemonset_overhead.yaml"
+    file_path.touch(exist_ok=True)
+
+    # acquire a Cluster object
+    config_file_path = find_absolute_path_to_cluster_file(cluster_name)
+    with open(config_file_path) as f:
+        cluster = Cluster(yaml.load(f), config_file_path.parent)
+
+    # auth and inspect cluster
+    with cluster.auth():
+        k8s_dist, k8s_version = get_k8s_distribution()
+        ds_overhead = get_daemon_sets_overhead()
+
+    # read
+    with open(file_path) as f:
+        info = yaml.load(f) or {}
+
+    # update
+    ds_overhead["k8s_version"] = k8s_version
+    info.setdefault(k8s_dist, {})[cluster_name] = ds_overhead
+
+    # write
+    with open(file_path, "w") as f:
+        yaml.dump(info, f)

deployer/commands/generate/resource_allocation/daemonset_overhead.yaml

@@ -0,0 +1,180 @@
+# This file contains generated information about cpu/memory requests made by
+# DaemonSets with running pods in our clusters. This information is relevant
+# when planning cpu/memory requests for other pods as the daemonsets requests
+# reduces the available allocatable capacity.
+#
+# The requests vary between cloud providers, clusters, and k8s versions for
+# reasons like:
+#
+# - Cloud providers' managed k8s provides different DaemonSets by default
+# - DaemonSets may be coupled to managed k8s features (calico-node)
+# - DaemonSets' requests may be coupled to managed k8s version (netd)
+# - DaemonSets may have a vertical autoscaler changing requests dynamically over
+#   time if needed (calico-node-vertical-autoscaler)
+# - We may deploy or change a DaemonSet's requests over time (support-cryptnono,
+#   support-prometheus-node-exporter)
+#
+# This file isn't updated by automation, but can easily be updated by manually
+# by running a command once for each cluster:
+#
+#     ls config/clusters | xargs -I {} deployer generate resource-allocation daemonset-overhead {}
+#
+gke:
+  2i2c:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,gke-metrics-agent,ip-masq-agent,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: binder-staging-dind,binder-staging-image-cleaner,continuous-image-puller,imagebuilding-demo-binderhub-service-docker-api,netd
+    cpu_requests: 342m
+    memory_requests: 566Mi
+    k8s_version: v1.26.5-gke.2100
+  2i2c-uk:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,gke-metrics-agent,ip-masq-agent,netd,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 344m
+    memory_requests: 596Mi
+    k8s_version: v1.27.4-gke.900
+  awi-ciroh:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,gke-metrics-agent,ip-masq-agent,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: netd
+    cpu_requests: 342m
+    memory_requests: 566Mi
+    k8s_version: v1.25.10-gke.2700
+  callysto:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,gke-metrics-agent,ip-masq-agent,netd,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 344m
+    memory_requests: 596Mi
+    k8s_version: v1.27.4-gke.900
+  catalystproject-latam:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,ip-masq-agent,netd,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 338m
+    memory_requests: 496Mi
+    k8s_version: v1.27.3-gke.100
+  cloudbank:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,gke-metrics-agent,ip-masq-agent,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: continuous-image-puller,continuous-image-puller,continuous-image-puller,netd
+    cpu_requests: 342m
+    memory_requests: 566Mi
+    k8s_version: v1.26.5-gke.2100
+  hhmi:
+    requesting_daemon_sets: fluentbit-gke,gke-metadata-server,netd,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 228m
+    memory_requests: 480Mi
+    k8s_version: v1.27.3-gke.100
+  leap:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,gke-metrics-agent,ip-masq-agent,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: netd
+    cpu_requests: 342m
+    memory_requests: 566Mi
+    k8s_version: v1.25.10-gke.2700
+  linked-earth:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,gke-metrics-agent,ip-masq-agent,netd,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 344m
+    memory_requests: 596Mi
+    k8s_version: v1.27.4-gke.900
+  m2lines:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,gke-metrics-agent,ip-masq-agent,netd,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 344m
+    memory_requests: 596Mi
+    k8s_version: v1.27.4-gke.900
+  meom-ige:
+    requesting_daemon_sets: fluentbit-gke,gke-metadata-server,gke-metrics-agent,netd,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 234m
+    memory_requests: 580Mi
+    k8s_version: v1.27.4-gke.900
+  pangeo-hubs:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,gke-metrics-agent,ip-masq-agent,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: netd
+    cpu_requests: 342m
+    memory_requests: 566Mi
+    k8s_version: v1.26.5-gke.2100
+  qcl:
+    requesting_daemon_sets: calico-node,fluentbit-gke,gke-metadata-server,ip-masq-agent,pdcsi-node,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: continuous-image-puller,continuous-image-puller,netd
+    cpu_requests: 336m
+    memory_requests: 466Mi
+    k8s_version: v1.25.10-gke.2700
+eks:
+  2i2c-aws-us:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.25.12-eks-2d98532
+  carbonplan:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.24.16-eks-2d98532
+  catalystproject-africa:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.27.4-eks-2d98532
+  gridsst:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.25.12-eks-2d98532
+  jupyter-meets-the-earth:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.25.12-eks-2d98532
+  nasa-cryo:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.25.12-eks-2d98532
+  nasa-ghg:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.27.4-eks-2d98532
+  nasa-veda:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.25.12-eks-2d98532
+  openscapes:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.24.16-eks-2d98532
+  smithsonian:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.25.12-eks-2d98532
+  ubc-eoas:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.24.17-eks-f8587cb
+  victor:
+    requesting_daemon_sets: aws-node,ebs-csi-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: ""
+    cpu_requests: 170m
+    memory_requests: 250Mi
+    k8s_version: v1.25.12-eks-2d98532
+aks:
+  utoronto:
+    requesting_daemon_sets: cloud-node-manager,csi-azuredisk-node,csi-azurefile-node,kube-proxy,support-cryptnono,support-prometheus-node-exporter
+    other_daemon_sets: calico-node,continuous-image-puller,continuous-image-puller,continuous-image-puller,continuous-image-puller
+    cpu_requests: 226m
+    memory_requests: 300Mi
+    k8s_version: v1.26.3