For the liveness probes, for example, this configuration means the server could take until failureThreshold*periodSeconds=30 seconds to get started. If such a condition doesn't occur the cluster will kill the pod to start over it again (restart policy is always).
Sometimes the pod could need more than 30 seconds to start, for example:
The user has registered other liveness checks in their application code and to satisfy those conditions is needed more time.
The management operations executed at runtime when the pod is started could also take some time to be applied, which delays the availability of the endpoint used by the default liveness probe.
The node where the pod is running is slow.
In such a case, users could be tempted to increase the initialDelaySeconds or the failure failureThreshold configuration, however increasing the initialDelaySeconds will delay a fixed value, which could compromise the expected fast response. Increasing the failureThreshold could also break the established conditions you have decided to check your application at runtime.
The alternative to those issues is to have the ability to configure a Startup Probe. The Startup Probe can be used to delay the initial checks done by the liveness and readiness probes, allowing some configuration without compromising them.
The goal of this RFE is to define the default initial configuration of a Startup Probe so users can tweak it when required.
Overview
The probes (Liveness and Readiness) added by the Helm Chart uses the following default values:
For the liveness probes, for example, this configuration means the server could take until
failureThreshold*periodSeconds=30 seconds
to get started. If such a condition doesn't occur the cluster will kill the pod to start over it again (restart policy is always).Sometimes the pod could need more than 30 seconds to start, for example:
In such a case, users could be tempted to increase the
initialDelaySeconds
or the failurefailureThreshold
configuration, however increasing theinitialDelaySeconds
will delay a fixed value, which could compromise the expected fast response. Increasing thefailureThreshold
could also break the established conditions you have decided to check your application at runtime.The alternative to those issues is to have the ability to configure a Startup Probe. The Startup Probe can be used to delay the initial checks done by the liveness and readiness probes, allowing some configuration without compromising them.
The goal of this RFE is to define the default initial configuration of a Startup Probe so users can tweak it when required.
Issue Metadata
https://issues.redhat.com/browse/EAP7-1769
Related Issues
Dev Contacts
Yeray Borgess yborgess@redhat.com
QE Contacts
TBD
Testing By
TBD
Affected Projects or Components
WildFly Charts
Other Interested Projects
N/A
Requirements
Hard Requirements
Nice-to-Have Requirements
N/A
Non-Requirements
N/A
Test Plan
Community Documentation
The Helm Chart docs will include a description of the uses of the Startup Probe.
Release Note Content
Added the ability to configure a Startup Probe.