public_git
/
postgres-operator
mirror of https://github.com/zalando/postgres-operator.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

make minimum limits boundaries configurable (#808) 

* make minimum limits boundaries configurable
* add e2e test
This commit is contained in:
Felix Kunde 2020-02-03 11:43:18 +01:00 committed by GitHub
parent fddaf0fb73
commit 1f0312a014
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
17 changed files with 175 additions and 93 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
charts/postgres-operator/crds/operatorconfigurations.yaml
View File

@ -179,6 +179,12 @@ spec:
default_memory_request:
type: string
pattern: '^(\d+(e\d+)?|\d+(\.\d+)?(e\d+)?[EPTGMK]i?)$'
min_cpu_limit:
type: string
pattern: '^(\d+m|\d+(\.\d{1,3})?)$'
min_memory_limit:
type: string
pattern: '^(\d+(e\d+)?|\d+(\.\d+)?(e\d+)?[EPTGMK]i?)$'
timeouts:
type: object
properties:

26
charts/postgres-operator/values-crd.yaml
View File

@ -115,13 +115,17 @@ configKubernetes:
# configure resource requests for the Postgres pods
configPostgresPodResources:
# CPU limits for the postgres containers
default_cpu_limit: "3"
# cpu request value for the postgres containers
default_cpu_limit: "1"
# CPU request value for the postgres containers
default_cpu_request: 100m
# memory limits for the postgres containers
default_memory_limit: 1Gi
default_memory_limit: 500Mi
# memory request value for the postgres containers
default_memory_request: 100Mi
# hard CPU minimum required to properly run a Postgres cluster
min_cpu_limit: 250m
# hard memory minimum required to properly run a Postgres cluster
min_memory_limit: 250Mi
# timeouts related to some operator actions
configTimeouts:
@ -251,7 +255,7 @@ configScalyr:
# CPU rquest value for the Scalyr sidecar
scalyr_cpu_request: 100m
# Memory limit value for the Scalyr sidecar
scalyr_memory_limit: 1Gi
scalyr_memory_limit: 500Mi
# Memory request value for the Scalyr sidecar
scalyr_memory_request: 50Mi
@ -272,13 +276,13 @@ serviceAccount:
priorityClassName: ""
resources: {}
# limits:
# cpu: 100m
# memory: 300Mi
# requests:
# cpu: 100m
# memory: 300Mi
resources:
limits:
cpu: 500m
memory: 500Mi
requests:
cpu: 100m
memory: 250Mi
# Affinity for pod assignment
# Ref: https://kubernetes.io/docs/concepts/configuration/assign-pod-node/#affinity-and-anti-affinity

24
charts/postgres-operator/values.yaml
View File

@ -108,13 +108,17 @@ configKubernetes:
# configure resource requests for the Postgres pods
configPostgresPodResources:
# CPU limits for the postgres containers
default_cpu_limit: "3"
# cpu request value for the postgres containers
default_cpu_limit: "1"
# CPU request value for the postgres containers
default_cpu_request: 100m
# memory limits for the postgres containers
default_memory_limit: 1Gi
default_memory_limit: 500Mi
# memory request value for the postgres containers
default_memory_request: 100Mi
# hard CPU minimum required to properly run a Postgres cluster
min_cpu_limit: 250m
# hard memory minimum required to properly run a Postgres cluster
min_memory_limit: 250Mi
# timeouts related to some operator actions
configTimeouts:
@ -248,13 +252,13 @@ serviceAccount:
priorityClassName: ""
resources: {}
# limits:
# cpu: 100m
# memory: 300Mi
# requests:
# cpu: 100m
# memory: 300Mi
resources:
limits:
cpu: 500m
memory: 500Mi
requests:
cpu: 100m
memory: 250Mi
# Affinity for pod assignment
# Ref: https://kubernetes.io/docs/concepts/configuration/assign-pod-node/#affinity-and-anti-affinity

14
docs/reference/operator_parameters.md
View File

@ -318,11 +318,19 @@ CRD-based configuration.
* **default_cpu_limit**
CPU limits for the Postgres containers, unless overridden by cluster-specific
settings. The default is `3`.
settings. The default is `1`.
* **default_memory_limit**
memory limits for the Postgres containers, unless overridden by cluster-specific
settings. The default is `1Gi`.
settings. The default is `500Mi`.
* **min_cpu_limit**
hard CPU minimum what we consider to be required to properly run Postgres
clusters with Patroni on Kubernetes. The default is `250m`.
* **min_memory_limit**
hard memory minimum what we consider to be required to properly run Postgres
clusters with Patroni on Kubernetes. The default is `250Mi`.
## Operator timeouts
@ -579,4 +587,4 @@ scalyr sidecar. In the CRD-based configuration they are grouped under the
CPU limit value for the Scalyr sidecar. The default is `1`.
* **scalyr_memory_limit**
Memory limit value for the Scalyr sidecar. The default is `1Gi`.
Memory limit value for the Scalyr sidecar. The default is `500Mi`.

10
docs/user.md
View File

@ -232,11 +232,11 @@ spec:
memory: 300Mi
```
The minimum limit to properly run the `postgresql` resource is `256m` for `cpu`
and `256Mi` for `memory`. If a lower value is set in the manifest the operator
will cancel ADD or UPDATE events on this resource with an error. If no
resources are defined in the manifest the operator will obtain the configured
[default requests](reference/operator_parameters.md#kubernetes-resource-requests).
The minimum limits to properly run the `postgresql` resource are configured to
`250m` for `cpu` and `250Mi` for `memory`. If a lower value is set in the
manifest the operator will raise the limits to the configured minimum values.
If no resources are defined in the manifest they will be obtained from the
configured [default requests](reference/operator_parameters.md#kubernetes-resource-requests).
## Use taints and tolerations for dedicated PostgreSQL nodes

68
e2e/tests/test_e2e.py
View File

@ -58,6 +58,57 @@ class EndToEndTestCase(unittest.TestCase):
k8s.create_with_kubectl("manifests/minimal-postgres-manifest.yaml")
k8s.wait_for_pod_start('spilo-role=master')
@timeout_decorator.timeout(TEST_TIMEOUT_SEC)
def test_min_resource_limits(self):
'''
Lower resource limits below configured minimum and let operator fix it
'''
k8s = self.k8s
cluster_label = 'version=acid-minimal-cluster'
_, failover_targets = k8s.get_pg_nodes(cluster_label)
# configure minimum boundaries for CPU and memory limits
minCPULimit = '250m'
minMemoryLimit = '250Mi'
patch_min_resource_limits = {
"data": {
"min_cpu_limit": minCPULimit,
"min_memory_limit": minMemoryLimit
}
}
k8s.update_config(patch_min_resource_limits)
# lower resource limits below minimum
pg_patch_resources = {
"spec": {
"resources": {
"requests": {
"cpu": "10m",
"memory": "50Mi"
},
"limits": {
"cpu": "200m",
"memory": "200Mi"
}
}
}
}
k8s.api.custom_objects_api.patch_namespaced_custom_object(
"acid.zalan.do", "v1", "default", "postgresqls", "acid-minimal-cluster", pg_patch_resources)
k8s.wait_for_master_failover(failover_targets)
pods = k8s.api.core_v1.list_namespaced_pod(
'default', label_selector='spilo-role=master,' + cluster_label).items
self.assert_master_is_unique()
masterPod = pods[0]
self.assertEqual(masterPod.spec.containers[0].resources.limits['cpu'], minCPULimit,
"Expected CPU limit {}, found {}"
.format(minCPULimit, masterPod.spec.containers[0].resources.limits['cpu']))
self.assertEqual(masterPod.spec.containers[0].resources.limits['memory'], minMemoryLimit,
"Expected memory limit {}, found {}"
.format(minMemoryLimit, masterPod.spec.containers[0].resources.limits['memory']))
@timeout_decorator.timeout(TEST_TIMEOUT_SEC)
def test_multi_namespace_support(self):
'''
@ -76,10 +127,9 @@ class EndToEndTestCase(unittest.TestCase):
@timeout_decorator.timeout(TEST_TIMEOUT_SEC)
def test_scaling(self):
"""
'''
Scale up from 2 to 3 and back to 2 pods by updating the Postgres manifest at runtime.
"""
'''
k8s = self.k8s
labels = "version=acid-minimal-cluster"
@ -93,9 +143,9 @@ class EndToEndTestCase(unittest.TestCase):
@timeout_decorator.timeout(TEST_TIMEOUT_SEC)
def test_taint_based_eviction(self):
"""
'''
Add taint "postgres=:NoExecute" to node with master. This must cause a failover.
"""
'''
k8s = self.k8s
cluster_label = 'version=acid-minimal-cluster'
@ -145,7 +195,7 @@ class EndToEndTestCase(unittest.TestCase):
@timeout_decorator.timeout(TEST_TIMEOUT_SEC)
def test_logical_backup_cron_job(self):
"""
'''
Ensure we can (a) create the cron job at user request for a specific PG cluster
(b) update the cluster-wide image for the logical backup pod
(c) delete the job at user request
@ -153,7 +203,7 @@ class EndToEndTestCase(unittest.TestCase):
Limitations:
(a) Does not run the actual batch job because there is no S3 mock to upload backups to
(b) Assumes 'acid-minimal-cluster' exists as defined in setUp
"""
'''
k8s = self.k8s
@ -208,10 +258,10 @@ class EndToEndTestCase(unittest.TestCase):
"Expected 0 logical backup jobs, found {}".format(len(jobs)))
def assert_master_is_unique(self, namespace='default', version="acid-minimal-cluster"):
"""
'''
Check that there is a single pod in the k8s cluster with the label "spilo-role=master"
To be called manually after operations that affect pods
"""
'''
k8s = self.k8s
labels = 'spilo-role=master,version=' + version

4
manifests/complete-postgres-manifest.yaml
View File

@ -42,8 +42,8 @@ spec:
cpu: 10m
memory: 100Mi
limits:
cpu: 300m
memory: 300Mi
cpu: 500m
memory: 500Mi
patroni:
initdb:
encoding: "UTF8"

6
manifests/configmap.yaml
View File

@ -15,9 +15,9 @@ data:
# custom_pod_annotations: "keya:valuea,keyb:valueb"
db_hosted_zone: db.example.com
debug_logging: "true"
# default_cpu_limit: "3"
# default_cpu_limit: "1"
# default_cpu_request: 100m
# default_memory_limit: 1Gi
# default_memory_limit: 500Mi
# default_memory_request: 100Mi
docker_image: registry.opensource.zalan.do/acid/spilo-cdp-12:1.6-p16
# enable_admin_role_for_users: "true"
@ -48,6 +48,8 @@ data:
# master_pod_move_timeout: 10m
# max_instances: "-1"
# min_instances: "-1"
# min_cpu_limit: 250m
# min_memory_limit: 250Mi
# node_readiness_label: ""
# oauth_token_secret_name: postgresql-operator
# pam_configuration: |

6
manifests/operatorconfiguration.crd.yaml
View File

@ -155,6 +155,12 @@ spec:
default_memory_request:
type: string
pattern: '^(\d+(e\d+)?|\d+(\.\d+)?(e\d+)?[EPTGMK]i?)$'
min_cpu_limit:
type: string
pattern: '^(\d+m|\d+(\.\d{1,3})?)$'
min_memory_limit:
type: string
pattern: '^(\d+(e\d+)?|\d+(\.\d+)?(e\d+)?[EPTGMK]i?)$'
timeouts:
type: object
properties:

4
manifests/postgres-operator.yaml
View File

@ -19,10 +19,10 @@ spec:
imagePullPolicy: IfNotPresent
resources:
requests:
cpu: 500m
cpu: 100m
memory: 250Mi
limits:
cpu: 2000m
cpu: 500m
memory: 500Mi
securityContext:
runAsUser: 1000

8
manifests/postgresql-operator-default-configuration.yaml
View File

@ -54,10 +54,12 @@ configuration:
# toleration: {}
# watched_namespace: ""
postgres_pod_resources:
default_cpu_limit: "3"
default_cpu_limit: "1"
default_cpu_request: 100m
default_memory_limit: 1Gi
default_memory_limit: 500Mi
default_memory_request: 100Mi
# min_cpu_limit: 250m
# min_memory_limit: 250Mi
timeouts:
pod_label_wait_timeout: 10m
pod_deletion_wait_timeout: 10m
@ -115,6 +117,6 @@ configuration:
scalyr_cpu_limit: "1"
scalyr_cpu_request: 100m
# scalyr_image: ""
scalyr_memory_limit: 1Gi
scalyr_memory_limit: 500Mi
scalyr_memory_request: 50Mi
# scalyr_server_url: ""

8
pkg/apis/acid.zalan.do/v1/crds.go
View File

@ -810,6 +810,14 @@ var OperatorConfigCRDResourceValidation = apiextv1beta1.CustomResourceValidation
Type: "string",
Pattern: "^(\\d+(e\\d+)?|\\d+(\\.\\d+)?(e\\d+)?[EPTGMK]i?)$",
},
"min_cpu_limit": {
Type: "string",
Pattern: "^(\\d+m|\\d+(\\.\\d{1,3})?)$",
},
"min_memory_limit": {
Type: "string",
Pattern: "^(\\d+(e\\d+)?|\\d+(\\.\\d+)?(e\\d+)?[EPTGMK]i?)$",
},
},
},
"timeouts": {

2
pkg/apis/acid.zalan.do/v1/operator_configuration_type.go
View File

@ -79,6 +79,8 @@ type PostgresPodResourcesDefaults struct {
DefaultMemoryRequest string `json:"default_memory_request,omitempty"`
DefaultCPULimit string `json:"default_cpu_limit,omitempty"`
DefaultMemoryLimit string `json:"default_memory_limit,omitempty"`
MinCPULimit string `json:"min_cpu_limit,omitempty"`
MinMemoryLimit string `json:"min_memory_limit,omitempty"`
}
// OperatorTimeouts defines the timeout of ResourceCheck, PodWait, ReadyWait

56
pkg/cluster/cluster.go
View File

@ -227,8 +227,8 @@ func (c *Cluster) Create() error {
c.setStatus(acidv1.ClusterStatusCreating)
if err = c.validateResources(&c.Spec); err != nil {
return fmt.Errorf("insufficient resource limits specified: %v", err)
if err = c.enforceMinResourceLimits(&c.Spec); err != nil {
return fmt.Errorf("could not enforce minimum resource limits: %v", err)
}
for _, role := range []PostgresRole{Master, Replica} {
@ -495,38 +495,38 @@ func compareResourcesAssumeFirstNotNil(a *v1.ResourceRequirements, b *v1.Resourc
}
func (c *Cluster) validateResources(spec *acidv1.PostgresSpec) error {
// setting limits too low can cause unnecessary evictions / OOM kills
const (
cpuMinLimit = "256m"
memoryMinLimit = "256Mi"
)
func (c *Cluster) enforceMinResourceLimits(spec *acidv1.PostgresSpec) error {
var (
isSmaller bool
err error
)
// setting limits too low can cause unnecessary evictions / OOM kills
minCPULimit := c.OpConfig.MinCPULimit
minMemoryLimit := c.OpConfig.MinMemoryLimit
cpuLimit := spec.Resources.ResourceLimits.CPU
if cpuLimit != "" {
isSmaller, err = util.IsSmallerQuantity(cpuLimit, cpuMinLimit)
isSmaller, err = util.IsSmallerQuantity(cpuLimit, minCPULimit)
if err != nil {
return fmt.Errorf("error validating CPU limit: %v", err)
return fmt.Errorf("could not compare defined CPU limit %s with configured minimum value %s: %v", cpuLimit, minCPULimit, err)
}
if isSmaller {
return fmt.Errorf("defined CPU limit %s is below required minimum %s to properly run postgresql resource", cpuLimit, cpuMinLimit)
c.logger.Warningf("defined CPU limit %s is below required minimum %s and will be set to it", cpuLimit, minCPULimit)
spec.Resources.ResourceLimits.CPU = minCPULimit
}
}
memoryLimit := spec.Resources.ResourceLimits.Memory
if memoryLimit != "" {
isSmaller, err = util.IsSmallerQuantity(memoryLimit, memoryMinLimit)
isSmaller, err = util.IsSmallerQuantity(memoryLimit, minMemoryLimit)
if err != nil {
return fmt.Errorf("error validating memory limit: %v", err)
return fmt.Errorf("could not compare defined memory limit %s with configured minimum value %s: %v", memoryLimit, minMemoryLimit, err)
}
if isSmaller {
return fmt.Errorf("defined memory limit %s is below required minimum %s to properly run postgresql resource", memoryLimit, memoryMinLimit)
c.logger.Warningf("defined memory limit %s is below required minimum %s and will be set to it", memoryLimit, minMemoryLimit)
spec.Resources.ResourceLimits.Memory = minMemoryLimit
}
}
@ -543,7 +543,6 @@ func (c *Cluster) Update(oldSpec, newSpec *acidv1.Postgresql) error {
c.mu.Lock()
defer c.mu.Unlock()
oldStatus := c.Status
c.setStatus(acidv1.ClusterStatusUpdating)
c.setSpec(newSpec)
@ -555,22 +554,6 @@ func (c *Cluster) Update(oldSpec, newSpec *acidv1.Postgresql) error {
}
}()
if err := c.validateResources(&newSpec.Spec); err != nil {
err = fmt.Errorf("insufficient resource limits specified: %v", err)
// cancel update only when (already too low) pod resources were edited
// if cluster was successfully running before the update, continue but log a warning
isCPULimitSmaller, err2 := util.IsSmallerQuantity(newSpec.Spec.Resources.ResourceLimits.CPU, oldSpec.Spec.Resources.ResourceLimits.CPU)
isMemoryLimitSmaller, err3 := util.IsSmallerQuantity(newSpec.Spec.Resources.ResourceLimits.Memory, oldSpec.Spec.Resources.ResourceLimits.Memory)
if oldStatus.Running() && !isCPULimitSmaller && !isMemoryLimitSmaller && err2 == nil && err3 == nil {
c.logger.Warning(err)
} else {
updateFailed = true
return err
}
}
if oldSpec.Spec.PgVersion != newSpec.Spec.PgVersion { // PG versions comparison
c.logger.Warningf("postgresql version change(%q -> %q) has no effect", oldSpec.Spec.PgVersion, newSpec.Spec.PgVersion)
//we need that hack to generate statefulset with the old version
@ -616,6 +599,12 @@ func (c *Cluster) Update(oldSpec, newSpec *acidv1.Postgresql) error {
// Statefulset
func() {
if err := c.enforceMinResourceLimits(&c.Spec); err != nil {
c.logger.Errorf("could not sync resources: %v", err)
updateFailed = true
return
}
oldSs, err := c.generateStatefulSet(&oldSpec.Spec)
if err != nil {
c.logger.Errorf("could not generate old statefulset spec: %v", err)
@ -623,6 +612,9 @@ func (c *Cluster) Update(oldSpec, newSpec *acidv1.Postgresql) error {
return
}
// update newSpec to for latter comparison with oldSpec
c.enforceMinResourceLimits(&newSpec.Spec)
newSs, err := c.generateStatefulSet(&newSpec.Spec)
if err != nil {
c.logger.Errorf("could not generate new statefulset spec: %v", err)

16
pkg/cluster/sync.go
View File

@ -23,7 +23,6 @@ func (c *Cluster) Sync(newSpec *acidv1.Postgresql) error {
c.mu.Lock()
defer c.mu.Unlock()
oldStatus := c.Status
c.setSpec(newSpec)
defer func() {
@ -35,16 +34,6 @@ func (c *Cluster) Sync(newSpec *acidv1.Postgresql) error {
}
}()
if err = c.validateResources(&c.Spec); err != nil {
err = fmt.Errorf("insufficient resource limits specified: %v", err)
if oldStatus.Running() {
c.logger.Warning(err)
err = nil
} else {
return err
}
}
if err = c.initUsers(); err != nil {
err = fmt.Errorf("could not init users: %v", err)
return err
@ -76,6 +65,11 @@ func (c *Cluster) Sync(newSpec *acidv1.Postgresql) error {
return err
}
if err = c.enforceMinResourceLimits(&c.Spec); err != nil {
err = fmt.Errorf("could not enforce minimum resource limits: %v", err)
return err
}
c.logger.Debugf("syncing statefulsets")
if err = c.syncStatefulSet(); err != nil {
if !k8sutil.ResourceAlreadyExists(err) {

2
pkg/controller/operator_config.go
View File

@ -75,6 +75,8 @@ func (c *Controller) importConfigurationFromCRD(fromCRD *acidv1.OperatorConfigur
result.DefaultMemoryRequest = fromCRD.PostgresPodResources.DefaultMemoryRequest
result.DefaultCPULimit = fromCRD.PostgresPodResources.DefaultCPULimit
result.DefaultMemoryLimit = fromCRD.PostgresPodResources.DefaultMemoryLimit
result.MinCPULimit = fromCRD.PostgresPodResources.MinCPULimit
result.MinMemoryLimit = fromCRD.PostgresPodResources.MinMemoryLimit
// timeout config
result.ResourceCheckInterval = time.Duration(fromCRD.Timeouts.ResourceCheckInterval)

8
pkg/util/config/config.go
View File

@ -37,8 +37,10 @@ type Resources struct {
PodToleration map[string]string `name:"toleration" default:""`
DefaultCPURequest string `name:"default_cpu_request" default:"100m"`
DefaultMemoryRequest string `name:"default_memory_request" default:"100Mi"`
DefaultCPULimit string `name:"default_cpu_limit" default:"3"`
DefaultMemoryLimit string `name:"default_memory_limit" default:"1Gi"`
DefaultCPULimit string `name:"default_cpu_limit" default:"1"`
DefaultMemoryLimit string `name:"default_memory_limit" default:"500Mi"`
MinCPULimit string `name:"min_cpu_limit" default:"250m"`
MinMemoryLimit string `name:"min_memory_limit" default:"250Mi"`
PodEnvironmentConfigMap string `name:"pod_environment_configmap" default:""`
NodeReadinessLabel map[string]string `name:"node_readiness_label" default:""`
MaxInstances int32 `name:"max_instances" default:"-1"`
@ -66,7 +68,7 @@ type Scalyr struct {
ScalyrCPURequest string `name:"scalyr_cpu_request" default:"100m"`
ScalyrMemoryRequest string `name:"scalyr_memory_request" default:"50Mi"`
ScalyrCPULimit string `name:"scalyr_cpu_limit" default:"1"`
ScalyrMemoryLimit string `name:"scalyr_memory_limit" default:"1Gi"`
ScalyrMemoryLimit string `name:"scalyr_memory_limit" default:"500Mi"`
}
// LogicalBackup defines configuration for logical backup