Re: [Condor-users] Condor losing stored credentials

[Todd Tannenbaum <tannenba@xxxxxxxxxxx>]

Michael O'Donnell wrote:
> If this is a win32 error code it means ERROR_NO_LOGON_SERVERS
>
> Condor Error codes
> https://condor-wiki.cs.wisc.edu/index.cgi/wiki?p=MagicNumbers
>
> Win32 Error codes
> http://msdn.microsoft.com/en-us/library/ms681381%28v=vs.85%29.aspx
>
> I will dig around some more today and see if I can find anything. 

Correct, it is an error code that Condor is propagating from Windows. 
Little bit more from Windows documentation :

//
// MessageId: ERROR_NO_LOGON_SERVERS (1311)
//
// MessageText:
//
//  There are currently no logon servers available to service the logon
//  request.
//

Are you using a domain - perhaps occasionally your domain controller is 
inaccessible or overloaded or some such?

The error in the schedd log you saw is when Condor attempts to validate 
the password by logging in the user and checking is it succeeded. 
Perhaps there is some limit these days to how many users can be logged 
into a given copy of Windows these days, for instance maybe Windows 7 
Ultimate has a different number of max logins vs Windows 7 Pro vs 
Windows Server 2008 etc etc.  What flavor of Windows are you using on 
your submit machine, and how many different users have jobs submitted at 
any one point (and/or how many different users may be own processes on 
the submit machine) ?

But I still don't understand why Condor would "forget" the stored 
password until you reset it - in a quick look around, I didn't see 
anywhere Condor removes a stored password beyond the obvious 
"condor_store_cred delete". I was thinking I might find someplace where 
Condor (foolishly?) may remove the stored password if it failed to 
login, thinking the password was bad. But I didn't do a super through 
analysis, just some quick looking around...

Curious, are you running the condor_credd, i.e. when you run 
"condor_config_val credd_host" on your submit machine what does it say? 
 Note that I am guessing you do NOT need to run a condor_credd since 
you are not using run-as-owner, and thus all the credentials could just 
be stored locally in the registry of the submit machine.

regards,
Todd