[jbosstools-issues] [JBoss JIRA] (JBIDE-24868) Server adapter: Switch off pod livenessProbe.periodSecond property and router timeout during debug session

[Roland Huß (JIRA)]

    [ https://issues.jboss.org/browse/JBIDE-24868?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455910#comment-13455910 ] 

Roland Huß edited comment on JBIDE-24868 at 8/30/17 5:01 AM:
-------------------------------------------------------------

Some remarks:

1) actually I haven't experienced that kind of timeout much (so when debugging with IntelliJ the connection seemed to stay open, maybe IntelliJ is sending some kind of ping, intentional or unintentional).
2) the router timeout is imo the biggest pita. The solution via port-forwarding is often not possible:
    * A Single Page App's UI doesn't use relative URL to access the backend for requests, so can't be changed to use the forwarded port
   * The application is composed via MicroServices which are aggregated on the route level (i.e. the router spans and defines the REST URL namespace), so for an application to work it might be mandatory to go over the route (e.g. our Syndesis app is example of this kind of application)

3) On a second thought, I don't think anymore that this is critical. The recommendation to choose to suspend only on the thread should cover 99% of all use cases, so changing the deployment config on the fly (and causing a redeployment for this) is too much effort for too less gain imo.

Maybe we could get at least an option to Minishift which can influence the router's global timeout value? For many developers this would be already a big help.


was (Author: rhuss):
Some remarks:

1) actually I haven't experienced that kind of timeout much (so when debugging with IntelliJ the connection seemed to stay open, maybe IntelliJ is sending some kind of ping, intentional or unintentional).
2) the router timeout is imo the biggest pita. The solution via port-forwarding is often not possible:
    * A Single Page App's UI doesn't use relative URL to access the backend for requests, so can't be changed to use the forwarded port
   * The application is composed via MicroServices which are aggregated on the route level (i.e. the routed spans and defines the REST URL namespace), so for an application to work it might be mandatory to go over the route (e.g. our Syndesis app is example of this kind of application)

3) On a second thought, I don't think anymore that this is critical. The recommendation to choose to suspend only on the thread should cover 99% of all use cases, so changing the deployment config on the fly (and causing a redeployment for this) is too much effort for too less gain imo.

Maybe we could get at least an option to Minishift which can influence the router's global timeout value? For many developers this would be already a big help.

> Server adapter: Switch off pod livenessProbe.periodSecond property and router timeout during debug session
> ----------------------------------------------------------------------------------------------------------
>
>                 Key: JBIDE-24868
>                 URL: https://issues.jboss.org/browse/JBIDE-24868
>             Project: Tools (JBoss Tools)
>          Issue Type: Enhancement
>          Components: openshift
>    Affects Versions: 4.5.0.Final
>            Reporter: Aurélien Pupier
>            Assignee: Andre Dietisheim
>              Labels: openshift_v3, server_adapter
>             Fix For: 4.5.x
>
>
> it will avoid to have "debug connections always killed after 30s staying in a breakpoint"
> see https://twitter.com/ro14nd/status/895886024387067904 for source of suggestion
> k8 documentations on the matter are here: https://kubernetes.io/docs/tasks/configure-pod-container/configure-liveness-readiness-probes/



--
This message was sent by Atlassian JIRA
(v7.2.3#72005)