Skip to content

Allow configurable resource requests and limits received by helm chart. #196

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 21 commits into from
Jul 25, 2024
Merged

Allow configurable resource requests and limits received by helm chart. #196

Show file tree
Hide file tree
Changes from 6 commits
Commits
Show all changes
21 commits
Select commit Hold shift + click to select a range
de7eab2
Added ability for operator to have resource values configured from he…
musa-asad Jul 17, 2024
63c2aae
Changed environmental variables to be specific values, get default fr…
musa-asad Jul 18, 2024
75cee1e
Fixed typo.
musa-asad Jul 18, 2024
3441437
Let users configure them individually.
musa-asad Jul 23, 2024
9ebd7d4
Allowed configurable languages.
musa-asad Jul 23, 2024
8aaa635
Assume default values from helm chart and updated tests.
musa-asad Jul 23, 2024
a5978d2
Added requests.
musa-asad Jul 25, 2024
28833f3
Updated tests.
musa-asad Jul 25, 2024
4ac6133
Regular parse and fixed default config.
musa-asad Jul 25, 2024
0ef0d33
Check languages.
musa-asad Jul 25, 2024
c49f688
Merge branch 'main' into init-resource
musa-asad Jul 25, 2024
de83a78
Debug
musa-asad Jul 25, 2024
7efb03f
Debug.
musa-asad Jul 25, 2024
42b2d0e
Reset.
musa-asad Jul 25, 2024
f464590
Increasing request to test.
musa-asad Jul 25, 2024
aea1763
Increasing request to test.
musa-asad Jul 25, 2024
ae8ee5a
Increasing limit to test.
musa-asad Jul 25, 2024
96b8454
Reset defaults for testing.
musa-asad Jul 25, 2024
2a6268c
Added defaults.
musa-asad Jul 25, 2024
42cf9b1
Addressed nits.
musa-asad Jul 25, 2024
9bdfa6c
Added link.
musa-asad Jul 25, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 29 additions & 12 deletions main.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -81,6 +81,11 @@ func stringFlagOrEnv(p *string, name string, envName string, defaultValue string
pflag.StringVar(p, name, defaultValue, usage)
}

func setEnvLang(lang string, cfg map[string]string) {
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_CPU_"+lang, cfg["cpu"])
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_MEMORY_"+lang, cfg["memory"])
}

func main() {
// registers any flags that underlying libraries might use
opts := zap.Options{}
Expand All @@ -93,23 +98,25 @@ func main() {

// add flags related to this operator
var (
metricsAddr string
probeAddr string
pprofAddr string
agentImage string
autoInstrumentationJava string
autoInstrumentationPython string
autoInstrumentationDotNet string
autoAnnotationConfigStr string
webhookPort int
tlsOpt tlsConfig
dcgmExporterImage string
neuronMonitorImage string
metricsAddr string
probeAddr string
pprofAddr string
autoInstrumentationConfigStr string
agentImage string
autoInstrumentationJava string
autoInstrumentationPython string
autoInstrumentationDotNet string
autoAnnotationConfigStr string
webhookPort int
tlsOpt tlsConfig
dcgmExporterImage string
neuronMonitorImage string
)

pflag.StringVar(&metricsAddr, "metrics-addr", ":8080", "The address the metric endpoint binds to.")
pflag.StringVar(&probeAddr, "health-probe-addr", ":8081", "The address the probe endpoint binds to.")
pflag.StringVar(&pprofAddr, "pprof-addr", "", "The address to expose the pprof server. Default is empty string which disables the pprof server.")
pflag.StringVar(&autoInstrumentationConfigStr, "auto-instrumentation-config", "", "The configuration for auto-instrumentation.")
stringFlagOrEnv(&agentImage, "agent-image", "RELATED_IMAGE_COLLECTOR", fmt.Sprintf("%s:%s", cloudwatchAgentImageRepository, v.AmazonCloudWatchAgent), "The default CloudWatch Agent image. This image is used when no image is specified in the CustomResource.")
stringFlagOrEnv(&autoInstrumentationJava, "auto-instrumentation-java-image", "RELATED_IMAGE_AUTO_INSTRUMENTATION_JAVA", fmt.Sprintf("%s:%s", autoInstrumentationJavaImageRepository, v.AutoInstrumentationJava), "The default OpenTelemetry Java instrumentation image. This image is used when no image is specified in the CustomResource.")
stringFlagOrEnv(&autoInstrumentationPython, "auto-instrumentation-python-image", "RELATED_IMAGE_AUTO_INSTRUMENTATION_PYTHON", fmt.Sprintf("%s:%s", autoInstrumentationPythonImageRepository, v.AutoInstrumentationPython), "The default OpenTelemetry Python instrumentation image. This image is used when no image is specified in the CustomResource.")
Expand All @@ -119,6 +126,16 @@ func main() {
stringFlagOrEnv(&neuronMonitorImage, "neuron-monitor-image", "RELATED_IMAGE_NEURON_MONITOR", fmt.Sprintf("%s:%s", neuronMonitorImageRepository, v.NeuronMonitor), "The default Neuron monitor image. This image is used when no image is specified in the CustomResource.")
pflag.Parse()

// set instrumentation cpu and memory limits in environment variables to be used for default instrumentation
autoInstrumentationConfig := map[string]map[string]string{"java": {"cpu": "500m", "memory": "64Mi"}, "python": {"cpu": "500m", "memory": "32Mi"}, "dotnet": {"cpu": "500m", "memory": "128Mi"}}
err := json.Unmarshal([]byte(autoInstrumentationConfigStr), &autoInstrumentationConfig)
if err != nil {
setupLog.Error(err, "Unable to unmarshal auto-instrumentation config, assuming default values")
}
setEnvLang("JAVA", autoInstrumentationConfig["java"])
setEnvLang("PYTHON", autoInstrumentationConfig["python"])
setEnvLang("DOTNET", autoInstrumentationConfig["dotnet"])

// set supported language instrumentation images in environment variable to be used for default instrumentation
os.Setenv("AUTO_INSTRUMENTATION_JAVA", autoInstrumentationJava)
os.Setenv("AUTO_INSTRUMENTATION_PYTHON", autoInstrumentationPython)
Expand Down
24 changes: 24 additions & 0 deletions pkg/instrumentation/defaultinstrumentation.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@ package instrumentation
import (
"errors"
"fmt"
"k8s.io/apimachinery/pkg/api/resource"
"os"

corev1 "k8s.io/api/core/v1"
Expand All @@ -25,6 +26,20 @@ const (
https = "https"
)

func getInstrumentationConfigLimits(lang string) corev1.ResourceList {
instrumentationConfigCpu, _ := os.LookupEnv("AUTO_INSTRUMENTATION_LIMIT_CPU_" + lang)
instrumentationConfigMemory, _ := os.LookupEnv("AUTO_INSTRUMENTATION_LIMIT_MEMORY_" + lang)

instrumentationConfigLimits := corev1.ResourceList{}
if instrumentationConfigCpu != "" {
instrumentationConfigLimits[corev1.ResourceCPU] = resource.MustParse(instrumentationConfigCpu)
}
if instrumentationConfigMemory != "" {
instrumentationConfigLimits[corev1.ResourceMemory] = resource.MustParse(instrumentationConfigMemory)
}
return instrumentationConfigLimits
}

func getDefaultInstrumentation(agentConfig *adapters.CwaConfig, isWindowsPod bool) (*v1alpha1.Instrumentation, error) {
javaInstrumentationImage, ok := os.LookupEnv("AUTO_INSTRUMENTATION_JAVA")
if !ok {
Expand Down Expand Up @@ -86,6 +101,9 @@ func getDefaultInstrumentation(agentConfig *adapters.CwaConfig, isWindowsPod boo
{Name: "OTEL_METRICS_EXPORTER", Value: "none"},
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
},
Resources: corev1.ResourceRequirements{
Limits: getInstrumentationConfigLimits("JAVA"),
},
},
Python: v1alpha1.Python{
Image: pythonInstrumentationImage,
Expand All @@ -103,6 +121,9 @@ func getDefaultInstrumentation(agentConfig *adapters.CwaConfig, isWindowsPod boo
{Name: "OTEL_PYTHON_CONFIGURATOR", Value: "aws_configurator"},
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
},
Resources: corev1.ResourceRequirements{
Limits: getInstrumentationConfigLimits("PYTHON"),
},
},
DotNet: v1alpha1.DotNet{
Image: dotNetInstrumentationImage,
Expand All @@ -120,6 +141,9 @@ func getDefaultInstrumentation(agentConfig *adapters.CwaConfig, isWindowsPod boo
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
{Name: "OTEL_DOTNET_AUTO_PLUGINS", Value: "AWS.Distro.OpenTelemetry.AutoInstrumentation.Plugin, AWS.Distro.OpenTelemetry.AutoInstrumentation"},
},
Resources: corev1.ResourceRequirements{
Limits: getInstrumentationConfigLimits("DOTNET"),
},
},
},
}, nil
Expand Down
85 changes: 85 additions & 0 deletions pkg/instrumentation/defaultinstrumentation_test.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@
package instrumentation

import (
"k8s.io/apimachinery/pkg/api/resource"
"os"
"reflect"
"testing"
Expand All @@ -19,6 +20,12 @@ func Test_getDefaultInstrumentationLinux(t *testing.T) {
os.Setenv("AUTO_INSTRUMENTATION_JAVA", defaultJavaInstrumentationImage)
os.Setenv("AUTO_INSTRUMENTATION_PYTHON", defaultPythonInstrumentationImage)
os.Setenv("AUTO_INSTRUMENTATION_DOTNET", defaultDotNetInstrumentationImage)
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_CPU_JAVA", "500m")
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_MEMORY_JAVA", "64Mi")
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_CPU_PYTHON", "500m")
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_MEMORY_PYTHON", "32Mi")
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_CPU_DOTNET", "500m")
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_MEMORY_DOTNET", "128Mi")

httpInst := &v1alpha1.Instrumentation{
Status: v1alpha1.InstrumentationStatus{},
Expand Down Expand Up @@ -51,6 +58,12 @@ func Test_getDefaultInstrumentationLinux(t *testing.T) {
{Name: "OTEL_METRICS_EXPORTER", Value: "none"},
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("64Mi"),
},
},
},
Python: v1alpha1.Python{
Image: defaultPythonInstrumentationImage,
Expand All @@ -68,6 +81,12 @@ func Test_getDefaultInstrumentationLinux(t *testing.T) {
{Name: "OTEL_PYTHON_CONFIGURATOR", Value: "aws_configurator"},
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("32Mi"),
},
},
},
DotNet: v1alpha1.DotNet{
Image: defaultDotNetInstrumentationImage,
Expand All @@ -85,6 +104,12 @@ func Test_getDefaultInstrumentationLinux(t *testing.T) {
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
{Name: "OTEL_DOTNET_AUTO_PLUGINS", Value: "AWS.Distro.OpenTelemetry.AutoInstrumentation.Plugin, AWS.Distro.OpenTelemetry.AutoInstrumentation"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("128Mi"),
},
},
},
},
}
Expand Down Expand Up @@ -119,6 +144,12 @@ func Test_getDefaultInstrumentationLinux(t *testing.T) {
{Name: "OTEL_METRICS_EXPORTER", Value: "none"},
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("64Mi"),
},
},
},
Python: v1alpha1.Python{
Image: defaultPythonInstrumentationImage,
Expand All @@ -136,6 +167,12 @@ func Test_getDefaultInstrumentationLinux(t *testing.T) {
{Name: "OTEL_PYTHON_CONFIGURATOR", Value: "aws_configurator"},
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("32Mi"),
},
},
},
DotNet: v1alpha1.DotNet{
Image: defaultDotNetInstrumentationImage,
Expand All @@ -153,6 +190,12 @@ func Test_getDefaultInstrumentationLinux(t *testing.T) {
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
{Name: "OTEL_DOTNET_AUTO_PLUGINS", Value: "AWS.Distro.OpenTelemetry.AutoInstrumentation.Plugin, AWS.Distro.OpenTelemetry.AutoInstrumentation"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("128Mi"),
},
},
},
},
}
Expand Down Expand Up @@ -218,6 +261,12 @@ func Test_getDefaultInstrumentationWindows(t *testing.T) {
os.Setenv("AUTO_INSTRUMENTATION_JAVA", defaultJavaInstrumentationImage)
os.Setenv("AUTO_INSTRUMENTATION_PYTHON", defaultPythonInstrumentationImage)
os.Setenv("AUTO_INSTRUMENTATION_DOTNET", defaultDotNetInstrumentationImage)
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_CPU_JAVA", "500m")
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_MEMORY_JAVA", "64Mi")
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_CPU_PYTHON", "500m")
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_MEMORY_PYTHON", "32Mi")
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_CPU_DOTNET", "500m")
os.Setenv("AUTO_INSTRUMENTATION_LIMIT_MEMORY_DOTNET", "128Mi")

httpInst := &v1alpha1.Instrumentation{
Status: v1alpha1.InstrumentationStatus{},
Expand Down Expand Up @@ -250,6 +299,12 @@ func Test_getDefaultInstrumentationWindows(t *testing.T) {
{Name: "OTEL_METRICS_EXPORTER", Value: "none"},
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("64Mi"),
},
},
},
Python: v1alpha1.Python{
Image: defaultPythonInstrumentationImage,
Expand All @@ -267,6 +322,12 @@ func Test_getDefaultInstrumentationWindows(t *testing.T) {
{Name: "OTEL_PYTHON_CONFIGURATOR", Value: "aws_configurator"},
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("32Mi"),
},
},
},
DotNet: v1alpha1.DotNet{
Image: defaultDotNetInstrumentationImage,
Expand All @@ -284,6 +345,12 @@ func Test_getDefaultInstrumentationWindows(t *testing.T) {
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
{Name: "OTEL_DOTNET_AUTO_PLUGINS", Value: "AWS.Distro.OpenTelemetry.AutoInstrumentation.Plugin, AWS.Distro.OpenTelemetry.AutoInstrumentation"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("128Mi"),
},
},
},
},
}
Expand Down Expand Up @@ -318,6 +385,12 @@ func Test_getDefaultInstrumentationWindows(t *testing.T) {
{Name: "OTEL_METRICS_EXPORTER", Value: "none"},
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("64Mi"),
},
},
},
Python: v1alpha1.Python{
Image: defaultPythonInstrumentationImage,
Expand All @@ -335,6 +408,12 @@ func Test_getDefaultInstrumentationWindows(t *testing.T) {
{Name: "OTEL_PYTHON_CONFIGURATOR", Value: "aws_configurator"},
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("32Mi"),
},
},
},
DotNet: v1alpha1.DotNet{
Image: defaultDotNetInstrumentationImage,
Expand All @@ -352,6 +431,12 @@ func Test_getDefaultInstrumentationWindows(t *testing.T) {
{Name: "OTEL_LOGS_EXPORTER", Value: "none"},
{Name: "OTEL_DOTNET_AUTO_PLUGINS", Value: "AWS.Distro.OpenTelemetry.AutoInstrumentation.Plugin, AWS.Distro.OpenTelemetry.AutoInstrumentation"},
},
Resources: corev1.ResourceRequirements{
Limits: corev1.ResourceList{
corev1.ResourceCPU: resource.MustParse("500m"),
corev1.ResourceMemory: resource.MustParse("128Mi"),
},
},
},
},
}
Expand Down
Loading