Users unable to create a new Che workspace in OpenShift.io if 10 users attempt to open Che workspaces in a ~2 minute window

[ldimaggi]

Related to redhat-developer/rh-che#629
Steps to recreate:

• Create a new space, new app/quickstart
• Attempt to create a new Che workspace
• The workspace fails to start - hangs on this screen:

NOTE: workaround - press "Start" button

No errors are seen in the Che6 log:

{"@timestamp":"2018-04-12T16:07:54.886+00:00","@version":1,"message":"Workspace
 'ldimaggi@redhat.com/april12-nyvn3' with id 'workspaceuimpi9knqfw8jkob' created by user
 'ldimaggi@redhat.com'","logger_name":"org.eclipse.che.api.workspace.server.WorkspaceManager"
,"thread_name":"http-nio-8080-exec-13","level":"INFO","level_value":20000,"req_id"
:"6ca0e5a5-077a-4134-aa0a-e524b7ea3db9","identity_id":"20ddc23a-bb62-4834-9130-9af2f54e85b1"}
``` #

[ldimaggi]

Related to the number of concurrent users in Che?

[ibuziuk]

Related to redhat-developer/rh-che#629

[ibuziuk]

Related to the number of concurrent users in Che?

might be, @rhopp is going to perform some load tesing AFAIK

[ghost]

@ibuziuk is it still Che starter that is in charge of staring workspaces? According to server log, the workspace has been created but never started. Auto start of workspaces is set to false as far as I understand. Hence, when IDE link is opened in a new tab, there's this message and the IDE does not auto start the workspace.

[ibuziuk]

@eivantsov yeah, che-starter is in charge of starting workpaces. It seems that the issue occurs under some load when dozen of users tries to start a workspace. Assuming that workspace startup is implemented asynchronously it might be a matter of configuring async thread pool executor on che-starter side.

However, if you take a look at description of rh-che issue redhat-developer/rh-che#629 the weird part there is The issue is getting problematic when the user start a second workspace: Che treats first workspace as "Started / Starting" even thought it is stopped - redhat-developer/rh-che#629 (comment)

WDYT about this one ? ^

Also, integration tests in che-starter caught some unexpected problems with workspace status flapping between running & starting - redhat-developer/rh-che#629 (comment)

one more thing is similar problem with dashboard eclipse-che/che#9381
(not sure if it related, but the context is quite the same)

[ldimaggi]

Are the parameters to force the error: 10 users - regardless of whether their accounts are configured on the same or different clusters - if they all attempt to start a workspace in the same ~2 minute window?

[l0rd]

@ldimaggi I am closing this one since it's a duplicate of #3135. Please re-open if you consider these as distinct issues.

[ldimaggi]

That's fine - if you believe that the same issue is described in both issues.