Upgrading SSO 5.5

*EDIT* After 3 hours on the phone with VMware Technical Support we are finally running again. Permissions on one vCenter were wiped and have to be recreated and a bug with Win2k12 domain controllers has hit is meaning that we had to create each domain as an identity source. But we are running again! *EDIT*


This is going to be a short post about my current experiences with upgrading the SSO component of vCenter. Short because it is still not working and I am waiting for VMware support to contact me.

So after VMworld I was pushing to upgrade to vSphere 5.5, atleast web client and SSO as this would solve a lot of our AD problems. So planning began and in November I revived my old test environment for the vSphere 5.1 upgrade. I dusted it off, got it running (more on that in a later post) and started the upgrade. The upgrade went fine – as such. After coming online again I could not enumerate users in SSO groups. I could search the AD via the new Integrated Windows Authentication Identity Source but not enumerate members of SSO groups. The web client would be “loading” for a LOOONG time and the following could be seen over and over again in the vmware-sts-idmd.log:

07:07:04,885 WARN   [LdapErrorChecker] Error received by LDAP client: com.vmware.identity.interop.ldap.WinLdapClientLibrary, error code: 81
07:07:04,885 WARN   [ServerUtils] cannot bind connection: [ldap://ADSERVERNAME, null]
07:07:04,885 ERROR  [ServerUtils] cannot establish connection with uri: [ldap://ADSERVERNAME]
07:07:04,885 ERROR  [ActiveDirectoryProvider] Failed to get GC connection to domain aau.dkLdap_sasl_bind failedServer Down

I purged dates and servernames from the log. But in essence it was looking like the SSO was attempting to make a Global Catalog (GC) connection to an AD server but it was selecting an AD server which was not running the GC service and when attempting to connect it decides that the server is down but keeps trying to contact it. Our AD forest consists of 1 root domain and about 20 subdomains and in total about 55 domain controllers where a bit over half run the GC service. There is no setting to tell SSO which server to talk to in the AD forest. So no change.

I got the following from VMware technical support a few days later:


1 Log in to the vSphere Web Client as administrator@vsphere.local or as another user with vCenter Single Sign-On administrator privileges.

2 Browse to Administration > Single Sign-On > Configuration.

3 On the Identity Sources tab, select an identity source and click the Set as Default Domain icon.

In the domain display, the default domain shows (default) in the Domain column.

Please restart the services and login to the web client/ vi client.

When I tried that it seemed to work. I reverted and it still worked leading me to think that it was not the above that fixed it but simply time as it would at some point select another server. I was assured over the phone by VMware technical support that customers that had seen the above error had fixed it with the above.

But now, to tell you the truth, the above log snippet is from today. I went ahead and upgrade SSO and Web client yesterday and got stuck with this error after spending about 2 hours reconnecting/repointing a vCenter 5.1 to the new lookup service. It requires you to change the vpxd.cfg file as it has hard coded the path to the STS service in that file and it points to the old /ims and not the new /sts url. It also caused a total wipe of permissions on the one vCenter I repointed causing some incosistent behavior in the web client. Looking under vCenter I cannot see it but if I browse down through Administration -> licensing and find it there I can see all objects with the admin@System-Domain account. The only account left with permissions on the vCenter.

I’m back at the office early this morning, slept like crap due to this. This is going to be a long day!