CF v237 - DEA - Restarting or restaging an app with a route mapped results in timeouts

Siva Balan <mailsiva@...>

After the upgrade to CF v237, we have started noticing a peculiar issue
with some of our apps(like UAA and custom service brokers). Below is the
summary of the issue(I will take UAA as an example where we see this issue):

1. When I try to restart or restage the UAA app with a route mapped to it,
it does not complete the startup. It seem to stop the startup in between
and waits for something(not sure what), and when the health manager is not
able to reach the app within a specified time period, it issues a restart
of the app. And this cycle continues.
2. When I try to restart or restage the UAA app with no routes bound to it,
it starts the app successfully but the logs still stall at the same place
as before. But after 3-4 mins, the logs begin to scroll and completes the
startup of the UAA server. The logs seem to stall right after the first
line below:

2016-07-20T12:16:41.00-0700 [App/0] OUT [Z: C:] [2016-07-20
19:16:41.004] cloudfoundry-identity-server - ???? [localhost-startStop-1]
.... INFO --- SamlLoginServerKeyManager: Loaded service provider
certificate service-provider-cert

The logging stalls here for a good 3-4 mins<<<<<<<

2016-07-20T12:19:55.69-0700 [App/0] OUT [Z: C:] [2016-07-20
19:19:55.696] cloudfoundry-identity-server - ???? [localhost-startStop-1]
.... INFO --- Audit: IdentityProviderModifiedEvent
originKey='keystone', name='keystone', type='keystone', active=false},
origin=[caller=null], identityZoneId=[uaa]

2016-07-20T12:19:55.72-0700 [App/0] OUT [Z: C:] [2016-07-20
19:19:55.722] cloudfoundry-identity-server - ???? [localhost-startStop-1]
.... INFO --- Audit: IdentityProviderModifiedEvent
originKey='ldap', name='ldap', type='ldap', active=false},
origin=[caller=null], identityZoneId=[uaa]

Its is as if the health check does not happen when no routes are mapped to
the UAA and hence it starts the app successfully, But with a route mapped
to the app, the health check fails and continues a cycle of app restarts.

We see this in not just the UAA app but also in some custom service broker
app. As mentioned in the subject, we are using DEAs and not on Diego.

2 questions:

1. Any idea what is causing this delay in starting up of app?
2. Does health check happen differently for an app when a route is not
mapped to it?

Any input is appreciated?




Join { to automatically receive all group messages.