Comment on page

Weakspots

Steadybit helps you in your path to a resilient system by analysing the configuration of your deployed environment and suggesting improvements to make it even more resilient. This way, you can fast-forward and don't need to execute an Experiment for learning your system.
We currently support some weak spots for Kubernetes, AWS, Azure and Java.
Kubernetes Weak Spots
Weak Spot
Description
Why is it important for a resilient system?
CPU Limit
Checks whether the consumption of the CPU is limited for a Kubernetes deployment.
When your deployment uses too much CPU other deployments on the same node could become unstable.
Memory Limit
Ensures that your configuration prevents a Kubernetes deployment to consume more memory than configured.
When your deployment uses too much memory other deployments on the same node could become unstable.
Single Replica
Verifies that at least two PODs of the same deployment exists.
It is hard to make single source of failure (like a single POD) resilient.
Liveness Probe
Ensures that a probe is configured so that Kubernetes is able to detect whether the POD is still working or it needs to be restarted.
An unresponsive POD may lead to no proper load balancing of requests. Sometimes, simple restart can help.
Readiness Probe
Tests the configuration of a rediness probe so that Kubernetes knows when a new starting POD is ready to receive traffic.
A POD that is still starting up may not be ready to accept traffic yet and thus lead to no proper load balancing of requests.
Readiness Probe minReady
Verifies whether Kubernetes is configured to prevent scheduling traffic immediately after the readinessProbe is successful.
May help to ensure that the newly started POD is stable enough to process traffic
Image Pull Policy
Checks that the imagePullPolicy is enforced by setting it to always to ensure that always the latest version is running.
If set otherwise the OCI runtime may execute an older cached version of the image. See also weakspot "image latest tag".
Image Latest Tag
Prevents usage of the latest tag as a version.
Using latest version makes it hard to identify which version is running and may also lead to different versions running in the same cluster.
Deployment Strategy
Enforces the deployment strategy rollingUpdate.
Prevents downtime while updating a deployment.
Horizontal POD Autoscaler
Verifies that a horizontal pod autoscaler is configured to schedule dynamically additional PODs.
Ensure that your system is properly scaled aligned to resource consumption.
POD Anti Affinity
Checks that your deployment is distributed across your cluster's nodes.
Just having multiple PODs (see "single replica") isn't enough in case all the PODs are running on a single node.
Single Node
Tests that your Kubernetes cluster is running on multiple nodes.
A Kubernetes cluster with a single node is hardly resilient against node outages.
AWS Kubernetes Weak Spots
Weak Spot
Description
Why is it important for a resilient system?
Single AWS Zone
Verifies that your Kubernetes cluster isn't running within a single Availability Zone.
A single Availability Zone is likely to fail which would lead to an unavailable Kubernetes cluster.
Azure Kubernetes Weak Spots
Weak Spot
Description
Why is it important for a resilient system?
Single Azure Zone
Verifies that your Kubernetes cluster isn't running within a single Availability Zone.
A single Availability Zone is likely to fail which would lead to an unavailable Kubernetes cluster.
JVM Weak Spots
Weak Spot
Description
Why is it important for a resilient system?
HTTP Call without Timeout
Ensures that your HTTP client is configured to have a timeout.
In case the target endpoint is unresponsive, the caller would wait forever on a response.
HTTP Call without Circuit Breaker
Verifies that your HTTP endpoint is configured with a circuit breaker.
Avoids cascading failures by short-circuiting to a configured fallback in case of problems.
Call for Feedback
Weak spots is a beta feature which we would like to get your opinion on. Feel free to reach out to us and share your experience.

Weak Spot	Description	Why is it important for a resilient system?
CPU Limit	Checks whether the consumption of the CPU is limited for a Kubernetes deployment.	When your deployment uses too much CPU other deployments on the same node could become unstable.
Memory Limit	Ensures that your configuration prevents a Kubernetes deployment to consume more memory than configured.	When your deployment uses too much memory other deployments on the same node could become unstable.
Single Replica	Verifies that at least two PODs of the same deployment exists.	It is hard to make single source of failure (like a single POD) resilient.
Liveness Probe	Ensures that a probe is configured so that Kubernetes is able to detect whether the POD is still working or it needs to be restarted.	An unresponsive POD may lead to no proper load balancing of requests. Sometimes, simple restart can help.
Readiness Probe	Tests the configuration of a rediness probe so that Kubernetes knows when a new starting POD is ready to receive traffic.	A POD that is still starting up may not be ready to accept traffic yet and thus lead to no proper load balancing of requests.
Readiness Probe minReady	Verifies whether Kubernetes is configured to prevent scheduling traffic immediately after the `readinessProbe` is successful.	May help to ensure that the newly started POD is stable enough to process traffic
Image Pull Policy	Checks that the `imagePullPolicy` is enforced by setting it to `always` to ensure that always the latest version is running.	If set otherwise the OCI runtime may execute an older cached version of the image. See also weakspot "image latest tag".
Image Latest Tag	Prevents usage of the `latest` tag as a version.	Using `latest` version makes it hard to identify which version is running and may also lead to different versions running in the same cluster.
Deployment Strategy	Enforces the deployment strategy `rollingUpdate`.	Prevents downtime while updating a deployment.
Horizontal POD Autoscaler	Verifies that a horizontal pod autoscaler is configured to schedule dynamically additional PODs.	Ensure that your system is properly scaled aligned to resource consumption.
POD Anti Affinity	Checks that your deployment is distributed across your cluster's nodes.	Just having multiple PODs (see "single replica") isn't enough in case all the PODs are running on a single node.
Single Node	Tests that your Kubernetes cluster is running on multiple nodes.	A Kubernetes cluster with a single node is hardly resilient against node outages.

Weak Spot	Description	Why is it important for a resilient system?
Single AWS Zone	Verifies that your Kubernetes cluster isn't running within a single Availability Zone.	A single Availability Zone is likely to fail which would lead to an unavailable Kubernetes cluster.

Weak Spot	Description	Why is it important for a resilient system?
Single Azure Zone	Verifies that your Kubernetes cluster isn't running within a single Availability Zone.	A single Availability Zone is likely to fail which would lead to an unavailable Kubernetes cluster.

Weak Spot	Description	Why is it important for a resilient system?
HTTP Call without Timeout	Ensures that your HTTP client is configured to have a timeout.	In case the target endpoint is unresponsive, the caller would wait forever on a response.
HTTP Call without Circuit Breaker	Verifies that your HTTP endpoint is configured with a circuit breaker.	Avoids cascading failures by short-circuiting to a configured fallback in case of problems.

Use Steadybit - Previous

Landscape

Next - Use Steadybit

Actions

Last modified 5mo ago