NoSuchFlowExecutionException with office365 apps

[Cantor, Scott]

On 6/26/17, 6:51 PM, "users on behalf of Klingenstein, Nate" <users-bounces at shibboleth.net on behalf of nklingenstein at calstate.edu> wrote:

> That error means that a user showed up with a Spring conversation that the node couldn't route.  I most frequently see it during
> node restarts here, but a 50% error rate strongly hints at something being off in your session replication or stickiness.

Lack of stickyness would certainly seem the most likely cause. The error simply means there was no flow conversation matching the "e2s1" parameter value in the URL, nothing deep about it. That's just how it recovers the state from memory and that's per-node state, it's not replicated or possible to replicate at this time.

This certainly has nothing to do with Office 365.

> I think these errors are virtually impossible to manually trace because state is such an abstract concept to the IdP, though others
> disagree.

It is about as far from abstract as it's possible to get.

With enough web logging, you can identify clients crossing nodes, and turning up SWF logging will show the execution keys it's creating or resuming, but "e2s1" just means "second conversation by the client using a Java session". The s1 is more difficult to explain but it just means that the first time the flow ends in a paused state at the client, the parameter it sends out is "s1" to reflect that it's the first time for that conversation.

-- Scott