zuul/nodepool
Code Issues Proposed changes

Allow custom k8s pod specs

Browse Source 
This change adds the ability to use the k8s (and friends) drivers
to create pods with custom specs.  This will allow nodepool admins
to define labels that create pods with options not otherwise supported
by Nodepool, as well as pods with multiple containers.

This can be used to implement the versatile sidecar pattern, which,
in a system where it is difficult to background a system process (such
as a database server or container runtime) is useful to run jobs with
such requirements.

It is still the case that a single resource is returned to Zuul, so
a single pod will be added to the inventory.  Therefore, the expectation
that it should be possible to shell into the first container in the
pod is documented.

Change-Id: I4a24a953a61239a8a52c9e7a2b68a7ec779f7a3d
This commit is contained in:
James E. Blair 2024-01-29 10:24:22 -08:00
parent e59bd1e331
commit c78fe769f2
18 changed files with 484 additions and 36 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
doc/source/kubernetes.rst
View File

@ -491,3 +491,41 @@ Selecting the kubernetes driver adds the following options to the
label type. Sets the `volumeMounts` flag on the
container. If supplied, this should be a list of
Kubernetes Container VolumeMount definitions.
.. attr:: spec
:type: dict
This attribute is exclusive with all other label
attributes except
:attr:`providers.[kubernetes].pools.labels.name`,
:attr:`providers.[kubernetes].pools.labels.type`,
:attr:`providers.[kubernetes].pools.labels.annotations`,
:attr:`providers.[kubernetes].pools.labels.labels` and
:attr:`providers.[kubernetes].pools.labels.dynamic-labels`.
If a `spec` is provided, then Nodepool will supply the
contents of this value verbatim to Kubernetes as the
``spec`` attribute of the Kubernetes ``Pod`` definition.
No other Nodepool attributes are used, including any
default values set at the provider level (such as
`default-label-cpu` and similar).
This attribute allows for the creation of arbitrary
complex pod definitions but the user is responsible for
ensuring that they are suitable. The first container in
the pod is expected to be a long-running container that
hosts a shell environment for running commands. The
following minimal definition matches what Nodepool itself
normally creates and is recommended as a starting point:
.. code-block:: yaml
labels:
- name: custom-pod
type: pod
spec:
containers:
- name: custom-pod
image: ubuntu:jammy
imagePullPolicy: IfNotPresent
command: ["/bin/sh", "-c"]
args: ["while true; do sleep 30; done;"]

36
doc/source/openshift-pods.rst
View File

@ -420,3 +420,39 @@ Selecting the openshift pods driver adds the following options to the
Sets the `volumeMounts` flag on the container. If
supplied, this should be a list of OpenShift Container
VolumeMount definitions.
.. attr:: spec
:type: dict
This attribute is exclusive with all other label
attributes except
:attr:`providers.[openshiftpods].pools.labels.name`
:attr:`providers.[openshiftpods].pools.labels.annotations`,
:attr:`providers.[openshiftpods].pools.labels.labels` and
:attr:`providers.[openshiftpods].pools.labels.dynamic-labels`.
If a `spec` is provided, then Nodepool will supply the
contents of this value verbatim to OpenShift as the
``spec`` attribute of the OpenShift ``Pod`` definition.
No other Nodepool attributes are used, including any
default values set at the provider level (such as
`default-label-cpu` and similar).
This attribute allows for the creation of arbitrary
complex pod definitions but the user is responsible for
ensuring that they are suitable. The first container in
the pod is expected to be a long-running container that
hosts a shell environment for running commands. The
following minimal definition matches what Nodepool itself
normally creates and is recommended as a starting point:
.. code-block:: yaml
labels:
- name: custom-pod
spec:
containers:
- name: custom-pod
image: ubuntu:jammy
imagePullPolicy: IfNotPresent
command: ["/bin/sh", "-c"]
args: ["while true; do sleep 30; done;"]

38
doc/source/openshift.rst
View File

@ -502,3 +502,41 @@ Selecting the openshift driver adds the following options to the
label type. Sets the `volumeMounts` flag on the
container. If supplied, this should be a list of
OpenShift Container VolumeMount definitions.
.. attr:: spec
:type: dict
This attribute is exclusive with all other label
attributes except
:attr:`providers.[openshift].pools.labels.name` and
:attr:`providers.[openshift].pools.labels.type`,
:attr:`providers.[openshift].pools.labels.annotations`,
:attr:`providers.[openshift].pools.labels.labels` and
:attr:`providers.[openshift].pools.labels.dynamic-labels`.
If a `spec` is provided, then Nodepool will supply the contents
of this value verbatim to OpenShift as the ``spec``
attribute of the OpenShift ``Pod`` definition. No other
Nodepool attributes are used, including any default values
set at the provider level (such as `default-label-cpu` and
similar).
This attribute allows for the creation of arbitrary
complex pod definitions but the user is responsible for
ensuring that they are suitable. The first container in
the pod is expected to be a long-running container that
hosts a shell environment for running commands. The
following minimal definition matches what Nodepool itself
normally creates and is recommended as a starting point:
.. code-block:: yaml
labels:
- name: custom-pod
type: pod
spec:
containers:
- name: custom-pod
image: ubuntu:jammy
imagePullPolicy: IfNotPresent
command: ["/bin/sh", "-c"]
args: ["while true; do sleep 30; done;"]

15
nodepool/driver/kubernetes/config.py
View File

@ -55,6 +55,7 @@ class KubernetesPool(ConfigPool):
pl = KubernetesLabel()
pl.name = label['name']
pl.type = label['type']
pl.spec = label.get('spec')
pl.image = label.get('image')
pl.image_pull = label.get('image-pull', 'IfNotPresent')
pl.python_path = label.get('python-path', 'auto')
@ -133,7 +134,7 @@ class KubernetesProviderConfig(ProviderConfig):
v.Required('value'): str,
}
k8s_label = {
k8s_label_from_nodepool = {
v.Required('name'): str,
v.Required('type'): str,
'image': str,
@ -160,10 +161,20 @@ class KubernetesProviderConfig(ProviderConfig):
'extra-resources': {str: int},
}
k8s_label_from_user = {
v.Required('name'): str,
v.Required('type'): str,
v.Required('spec'): dict,
'labels': dict,
'dynamic-labels': dict,
'annotations': dict,
}
pool = ConfigPool.getCommonSchemaDict()
pool.update({
v.Required('name'): str,
v.Required('labels'): [k8s_label],
v.Required('labels'): [v.Any(k8s_label_from_nodepool,
k8s_label_from_user)],
'max-cores': int,
'max-ram': int,
'max-resources': {str: int},

81
nodepool/driver/kubernetes/provider.py
View File

@ -304,6 +304,50 @@ class KubernetesProvider(Provider, QuotaSupport):
return resource
def createPod(self, node, pool, label, request):
if label.spec:
pod_body = self.getPodBodyCustom(node, pool, label, request)
else:
pod_body = self.getPodBodyNodepool(node, pool, label, request)
resource = self.createNamespace(node, pool, label, request,
restricted_access=True)
namespace = resource['namespace']
self.k8s_client.create_namespaced_pod(namespace, pod_body)
for retry in range(300):
pod = self.k8s_client.read_namespaced_pod(label.name, namespace)
if pod.status.phase == "Running":
break
self.log.debug("%s: pod status is %s", namespace, pod.status.phase)
time.sleep(1)
if retry == 299:
raise exceptions.LaunchNodepoolException(
"%s: pod failed to initialize (%s)" % (
namespace, pod.status.phase))
resource["pod"] = label.name
node.host_id = pod.spec.node_name
return resource
def getPodBodyCustom(self, node, pool, label, request):
k8s_labels = self._getK8sLabels(label, node, pool, request)
k8s_annotations = {}
if label.annotations:
k8s_annotations.update(label.annotations)
pod_body = {
'apiVersion': 'v1',
'kind': 'Pod',
'metadata': {
'name': label.name,
'labels': k8s_labels,
'annotations': k8s_annotations,
},
'spec': label.spec,
'restartPolicy': 'Never',
}
return pod_body
def getPodBodyNodepool(self, node, pool, label, request):
container_body = {
'name': label.name,
'image': label.image,
@ -335,8 +379,15 @@ class KubernetesProvider(Provider, QuotaSupport):
resources['requests'] = requests
if limits:
resources['limits'] = limits
if resources:
container_body['resources'] = resources
if label.volume_mounts:
container_body['volumeMounts'] = label.volume_mounts
if label.privileged is not None:
container_body['securityContext'] = {
'privileged': label.privileged,
}
spec_body = {
'containers': [container_body]
@ -344,21 +395,11 @@ class KubernetesProvider(Provider, QuotaSupport):
if label.node_selector:
spec_body['nodeSelector'] = label.node_selector
if label.scheduler_name:
spec_body['schedulerName'] = label.scheduler_name
if label.volumes:
spec_body['volumes'] = label.volumes
if label.volume_mounts:
container_body['volumeMounts'] = label.volume_mounts
if label.privileged is not None:
container_body['securityContext'] = {
'privileged': label.privileged,
}
k8s_labels = self._getK8sLabels(label, node, pool, request)
k8s_annotations = {}
if label.annotations:
@ -376,25 +417,7 @@ class KubernetesProvider(Provider, QuotaSupport):
'restartPolicy': 'Never',
}
resource = self.createNamespace(node, pool, label, request,
restricted_access=True)
namespace = resource['namespace']
self.k8s_client.create_namespaced_pod(namespace, pod_body)
for retry in range(300):
pod = self.k8s_client.read_namespaced_pod(label.name, namespace)
if pod.status.phase == "Running":
break
self.log.debug("%s: pod status is %s", namespace, pod.status.phase)
time.sleep(1)
if retry == 299:
raise exceptions.LaunchNodepoolException(
"%s: pod failed to initialize (%s)" % (
namespace, pod.status.phase))
resource["pod"] = label.name
node.host_id = pod.spec.node_name
return resource
return pod_body
def getRequestHandler(self, poolworker, request):
return handler.KubernetesNodeRequestHandler(poolworker, request)

15
nodepool/driver/openshift/config.py
View File

@ -57,6 +57,7 @@ class OpenshiftPool(ConfigPool):
pl = OpenshiftLabel()
pl.name = label['name']
pl.type = label['type']
pl.spec = label.get('spec')
pl.image = label.get('image')
pl.image_pull = label.get('image-pull', 'IfNotPresent')
pl.image_pull_secrets = label.get('image-pull-secrets', [])
@ -140,7 +141,7 @@ class OpenshiftProviderConfig(ProviderConfig):
v.Required('value'): str,
}
openshift_label = {
openshift_label_from_nodepool = {
v.Required('name'): str,
v.Required('type'): str,
'image': str,
@ -168,10 +169,20 @@ class OpenshiftProviderConfig(ProviderConfig):
'extra-resources': {str: int},
}
openshift_label_from_user = {
v.Required('name'): str,
v.Required('type'): str,
v.Required('spec'): dict,
'labels': dict,
'dynamic-labels': dict,
'annotations': dict,
}
pool = ConfigPool.getCommonSchemaDict()
pool.update({
v.Required('name'): str,
v.Required('labels'): [openshift_label],
v.Required('labels'): [v.Any(openshift_label_from_nodepool,
openshift_label_from_user)],
'max-cores': int,
'max-ram': int,
'max-resources': {str: int},

32
nodepool/driver/openshift/provider.py
View File

@ -222,6 +222,36 @@ class OpenshiftProvider(Provider, QuotaSupport):
def createPod(self, node, pool, project, pod_name, label, request):
self.log.debug("%s: creating pod in project %s" % (pod_name, project))
if label.spec:
pod_body = self.getPodBodyCustom(node, pool, pod_name, label,
request)
else:
pod_body = self.getPodBodyNodepool(node, pool, pod_name, label,
request)
self.k8s_client.create_namespaced_pod(project, pod_body)
def getPodBodyCustom(self, node, pool, pod_name, label, request):
k8s_labels = self._getK8sLabels(label, node, pool, request)
k8s_annotations = {}
if label.annotations:
k8s_annotations.update(label.annotations)
pod_body = {
'apiVersion': 'v1',
'kind': 'Pod',
'metadata': {
'name': pod_name,
'labels': k8s_labels,
'annotations': k8s_annotations,
},
'spec': label.spec,
'restartPolicy': 'Never',
}
return pod_body
def getPodBodyNodepool(self, node, pool, pod_name, label, request):
container_body = {
'name': label.name,
'image': label.image,
@ -296,7 +326,7 @@ class OpenshiftProvider(Provider, QuotaSupport):
'restartPolicy': 'Never',
}
self.k8s_client.create_namespaced_pod(project, pod_body)
return pod_body
def waitForPod(self, project, pod_name):
for retry in range(300):

13
nodepool/driver/openshiftpods/config.py
View File

@ -61,7 +61,7 @@ class OpenshiftPodsProviderConfig(OpenshiftProviderConfig):
v.Required('value'): str,
}
openshift_label = {
openshift_label_from_nodepool = {
v.Required('name'): str,
v.Required('image'): str,
'image-pull': str,
@ -88,10 +88,19 @@ class OpenshiftPodsProviderConfig(OpenshiftProviderConfig):
'extra-resources': {str: int},
}
openshift_label_from_user = {
v.Required('name'): str,
v.Required('spec'): dict,
'labels': dict,
'dynamic-labels': dict,
'annotations': dict,
}
pool = ConfigPool.getCommonSchemaDict()
pool.update({
v.Required('name'): str,
v.Required('labels'): [openshift_label],
v.Required('labels'): [v.Any(openshift_label_from_nodepool,
openshift_label_from_user)],
'max-cores': int,
'max-ram': int,
'max-resources': {str: int},

5
nodepool/tests/fixtures/config_validate/good.yaml vendored
View File

@ -21,6 +21,7 @@ labels:
- name: trusty-static
- name: kubernetes-namespace
- name: pod-fedora
- name: pod-custom
- name: openshift-project
- name: openshift-pod
- name: centos-ami
@ -145,6 +146,10 @@ providers:
labels:
- name: kubernetes-namespace
type: namespace
- name: pod-custom
type: pod
spec:
k8s: goes here
- name: pod-fedora
type: pod
image: docker.io/fedora:28

43
nodepool/tests/fixtures/config_validate/k8s_spec_duplicate.yaml vendored Normal file
View File

@ -0,0 +1,43 @@
elements-dir: /etc/nodepool/elements
images-dir: /opt/nodepool_dib
zookeeper-servers:
- host: zk1.openstack.org
port: 2181
chroot: /test
labels:
- name: pod-fedora
providers:
- name: kubespray
driver: kubernetes
context: admin-cluster.local
pools:
- name: main
labels:
- name: pod-fedora
type: pod
spec:
k8s: goes here
image: docker.io/fedora:28
cpu: 2
memory: 512
env:
- name: FOO
value: hello
- name: BAR
value: world
node-selector:
storageType: ssd
privileged: true
volumes:
- name: my-csi-inline-vol
csi:
driver: inline.storage.kubernetes.io
volume-mounts:
- mountPath: "/data"
name: my-csi-inline-vol
scheduler-name: niftyScheduler
labels:
environment: qa

10
nodepool/tests/fixtures/kubernetes.yaml vendored
View File

@ -15,6 +15,7 @@ tenant-resource-limits:
labels:
- name: pod-fedora
- name: pod-extra
- name: pod-custom
- name: kubernetes-namespace
providers:
@ -53,3 +54,12 @@ providers:
volume-mounts:
- name: my-csi-inline-vol
mountPath: /data
- name: pod-custom
type: pod
spec:
containers:
- name: pod-custom
image: ubuntu:jammy
imagePullPolicy: IfNotPresent
command: ["/bin/sh", "-c"]
args: ["while true; do sleep 30; done;"]

10
nodepool/tests/fixtures/openshift.yaml vendored
View File

@ -15,6 +15,7 @@ tenant-resource-limits:
labels:
- name: pod-fedora
- name: pod-extra
- name: pod-custom
- name: openshift-project
- name: pod-fedora-secret
@ -58,3 +59,12 @@ providers:
volume-mounts:
- name: my-csi-inline-vol
mountPath: /data
- name: pod-custom
type: pod
spec:
containers:
- name: pod-custom
image: ubuntu:jammy
imagePullPolicy: IfNotPresent
command: ["/bin/sh", "-c"]
args: ["while true; do sleep 30; done;"]

9
nodepool/tests/fixtures/openshiftpods.yaml vendored
View File

@ -15,6 +15,7 @@ tenant-resource-limits:
labels:
- name: pod-fedora
- name: pod-fedora-secret
- name: pod-custom
providers:
- name: openshift
@ -32,3 +33,11 @@ providers:
image: docker.io/fedora:28
image-pull-secrets:
- name: registry-secret
- name: pod-custom
spec:
containers:
- name: pod-custom
image: ubuntu:jammy
imagePullPolicy: IfNotPresent
command: ["/bin/sh", "-c"]
args: ["while true; do sleep 30; done;"]

9
nodepool/tests/unit/test_config_validator.py
View File

@ -93,3 +93,12 @@ class TestConfigValidation(tests.BaseTestCase):
validator = ConfigValidator(config)
ret = validator.validate()
self.assertEqual(ret, 1)
def test_k8s_spec_duplicate(self):
config = os.path.join(os.path.dirname(tests.__file__),
'fixtures', 'config_validate',
'k8s_spec_duplicate.yaml')
validator = ConfigValidator(config)
ret = validator.validate()
self.assertEqual(ret, 1)

52
nodepool/tests/unit/test_driver_kubernetes.py
View File

@ -611,3 +611,55 @@ class TestDriverKubernetes(tests.DBTestCase):
'public_ipv4': None,
'connection_port': 'redacted',
'state': 'ready'}, objs[0])
def test_kubernetes_custom(self):
# Test a pod with a custom spec
configfile = self.setup_config('kubernetes.yaml')
pool = self.useNodepool(configfile, watermark_sleep=1)
self.startPool(pool)
req = zk.NodeRequest()
req.state = zk.REQUESTED
req.tenant_name = 'tenant-1'
req.node_types.append('pod-custom')
self.zk.storeNodeRequest(req)
self.log.debug("Waiting for request %s", req.id)
req = self.waitForNodeRequest(req)
self.assertEqual(req.state, zk.FULFILLED)
self.assertNotEqual(req.nodes, [])
node = self.zk.getNode(req.nodes[0])
self.assertEqual(node.allocated_to, req.id)
self.assertEqual(node.state, zk.READY)
self.assertIsNotNone(node.launcher)
self.assertEqual(node.connection_type, 'kubectl')
self.assertEqual(node.connection_port.get('token'), 'fake-token')
self.assertEqual(node.attributes,
{'key1': 'value1', 'key2': 'value2'})
self.assertEqual(node.cloud, 'admin-cluster.local')
self.assertEqual(node.host_id, 'k8s-default-pool-abcd-1234')
ns, pod = self.fake_k8s_client._pod_requests[0]
self.assertEqual(pod['metadata'], {
'name': 'pod-custom',
'annotations': {},
'labels': {
'nodepool_node_id': '0000000000',
'nodepool_provider_name': 'kubespray',
'nodepool_pool_name': 'main',
'nodepool_node_label': 'pod-custom'
},
})
self.assertEqual(pod['spec'], {
'containers': [{
'name': 'pod-custom',
'image': 'ubuntu:jammy',
'imagePullPolicy': 'IfNotPresent',
'command': ['/bin/sh', '-c'],
'args': ['while true; do sleep 30; done;'],
}],
})
node.state = zk.DELETING
self.zk.storeNode(node)
self.waitForNodeDeletion(node)

52
nodepool/tests/unit/test_driver_openshift.py
View File

@ -617,3 +617,55 @@ class TestDriverOpenshift(tests.DBTestCase):
def test_openshift_tenant_quota_extra(self):
self._test_openshift_quota(
'openshift-tenant-quota-extra.yaml', pause=False)
def test_openshift_custom(self):
# Test a pod with a custom spec
configfile = self.setup_config('openshift.yaml')
pool = self.useNodepool(configfile, watermark_sleep=1)
self.startPool(pool)
req = zk.NodeRequest()
req.state = zk.REQUESTED
req.tenant_name = 'tenant-1'
req.node_types.append('pod-custom')
self.zk.storeNodeRequest(req)
self.log.debug("Waiting for request %s", req.id)
req = self.waitForNodeRequest(req)
self.assertEqual(req.state, zk.FULFILLED)
self.assertNotEqual(req.nodes, [])
node = self.zk.getNode(req.nodes[0])
self.assertEqual(node.allocated_to, req.id)
self.assertEqual(node.state, zk.READY)
self.assertIsNotNone(node.launcher)
self.assertEqual(node.connection_type, 'kubectl')
self.assertEqual(node.connection_port.get('token'), 'fake-token')
self.assertEqual(node.attributes,
{'key1': 'value1', 'key2': 'value2'})
self.assertEqual(node.cloud, 'admin-cluster.local')
self.assertIsNone(node.host_id)
ns, pod = self.fake_k8s_client._pod_requests[0]
self.assertEqual(pod['metadata'], {
'name': 'pod-custom',
'annotations': {},
'labels': {
'nodepool_node_id': '0000000000',
'nodepool_provider_name': 'openshift',
'nodepool_pool_name': 'main',
'nodepool_node_label': 'pod-custom'
},
})
self.assertEqual(pod['spec'], {
'containers': [{
'name': 'pod-custom',
'image': 'ubuntu:jammy',
'imagePullPolicy': 'IfNotPresent',
'command': ['/bin/sh', '-c'],
'args': ['while true; do sleep 30; done;'],
}],
})
node.state = zk.DELETING
self.zk.storeNode(node)
self.waitForNodeDeletion(node)

52
nodepool/tests/unit/test_driver_openshiftpods.py
View File

@ -441,3 +441,55 @@ class TestDriverOpenshiftPods(tests.DBTestCase):
'memory': '1024Mi'
},
})
def test_openshift_pod_custom(self):
# Test a pod with a custom spec
configfile = self.setup_config('openshiftpods.yaml')
pool = self.useNodepool(configfile, watermark_sleep=1)
self.startPool(pool)
req = zk.NodeRequest()
req.state = zk.REQUESTED
req.tenant_name = 'tenant-1'
req.node_types.append('pod-custom')
self.zk.storeNodeRequest(req)
self.log.debug("Waiting for request %s", req.id)
req = self.waitForNodeRequest(req)
self.assertEqual(req.state, zk.FULFILLED)
self.assertNotEqual(req.nodes, [])
node = self.zk.getNode(req.nodes[0])
self.assertEqual(node.allocated_to, req.id)
self.assertEqual(node.state, zk.READY)
self.assertIsNotNone(node.launcher)
self.assertEqual(node.connection_type, 'kubectl')
self.assertEqual(node.connection_port.get('token'), 'fake-token')
self.assertEqual(node.attributes,
{'key1': 'value1', 'key2': 'value2'})
self.assertEqual(node.cloud, 'service-account.local')
self.assertEqual(node.host_id, 'k8s-default-pool-abcd-1234')
ns, pod = self.fake_k8s_client._pod_requests[0]
self.assertEqual(pod['metadata'], {
'name': 'pod-custom-0000000000',
'annotations': {},
'labels': {
'nodepool_node_id': '0000000000',
'nodepool_provider_name': 'openshift',
'nodepool_pool_name': 'main',
'nodepool_node_label': 'pod-custom'
},
})
self.assertEqual(pod['spec'], {
'containers': [{
'name': 'pod-custom',
'image': 'ubuntu:jammy',
'imagePullPolicy': 'IfNotPresent',
'command': ['/bin/sh', '-c'],
'args': ['while true; do sleep 30; done;'],
}],
})
node.state = zk.DELETING
self.zk.storeNode(node)
self.waitForNodeDeletion(node)

10
releasenotes/notes/pod-spec-f590f3bd7852e8f6.yaml Normal file
View File

@ -0,0 +1,10 @@
---
features:
- |
It is now possible to provide an custom pod definition for the
Kubernetes and OpenShift drivers using the
:attr:`providers.[kubernetes].pools.labels.spec`,
:attr:`providers.[openshift].pools.labels.spec` and
:attr:`providers.[openshiftpods].pools.labels.spec` attributes. These
can be used to supply parameters not otherwise supported by
Nodepool, or to create complex pods with multiple containers.