error configuring Ops Manager authentication when using EPMC

[p-alexisv]

Following the Lab Guide to install PKSv1.5 using the Enterprise PKS Management Console (EPMC), one can hit a particular error during the Ops Manager deployment operation. When this error is hit in EPMC, the state of Deploy ops-manager 2.x operation is Failed, and the status is:

error configuring Ops Manager authentication: could not 
determine initial configuration status: Unexpected response 
code: 502 Bad Gateway

The 502 Bad Gateway error usually means that the Ops Manager backend services (web servers, UAA, etc.) are not yet fully ready when the UI was accessed. The services take some time to fully come up more especially in slower machines. It is likely that EPMC has exhausted all retries and exited, hence the error in EPMC.

To recover and resume the installation via EPMC, the following steps could be carried out:

1. check if the Ops Manager UI is up by manually loading the Ops Manager IP address URL in the Chrome Browser
2. if the Ops Manager page that loaded is the form where you can configure authentication, then that means you can safely resume the installation from EPMC. Otherwise, you may have to wait a liitle longer.
3. in EPMC, go to PKS Configuration, click Generate Configuration, and then click Apply Configuration to resume the installation.

[afewellvmware]

thanks for sharing all these details @p-alexisv I believe as you noted this may be a sporadic issue based on the underlying performance of a given running vPod. If this becomes a common issue, we may be able to tune the vApp some, but I dont think that will be a productive pursuit, as in occasions where there are performance issues with a given onecloud org, there are almost always other problems. I think the ideal resolution will be to ensure we have and can provide access to onecloud orgs that are not excessively oversubscribed and can provide consistent performance. The problem in this case is not with onecloud, but with the specific datacenter we are using is way oversubscribed, we are hoping to get new data center resources and to setup an environment we can prevent from oversubscription to provide more consistent performance. Some people today have access to really solid onecloud datacenters and everything works great, but we currently dont have enough capacity to share these broadly and so the orgs that most people can access have intermittent performance.

In any case I will leave this ticket open in case people run into this. The long-term resolution needs to be access to a more consistently performant backend, which will hopefully happen soon pending approval. Worse case we should have EPMC in the HOL pod by the next refresh cycle within a few months which should be very consistently reliable.