Chris,
I have confirmed that entering an Instance name of > 50 characters does indeed cause a: 500 Internal Server Error. So I am going to assume this is the issue.
This needs to be addressed both in Condor and in the RHEVM Driver. The Driver should do any validation checks before sending the API request to the BackEnd provider, and thus return an Error message with a more helpful message.
Cheers
Martyn
----- Original Message ----- From: "Chris Lalancette" clalance@redhat.com To: "Martyn Taylor" mtaylor@redhat.com Cc: "Chris Lalancette" clalance@redhat.com, "Michael Orazi" morazi@redhat.com Sent: Monday, 23 May, 2011 1:17:04 PM Subject: Re: e2e failing w/rhevm 2.2
On 05/23/11 - 07:38:12AM, Martyn Taylor wrote:
Hi Chris,
I've been trying this morning to get full e2e working with rhevm 2.2.
However, I'm still getting: Create_Instance_Failure: 500 Internal Server Error
You mentioned on Fridays call that there still might be some issues with Condor. Is this now fixed? Could this issue I'm seeing here?
Snippet of GridManager Log:
05/23/11 12:30:49 [18785] querying for removed/held jobs 05/23/11 12:30:49 [18785] Using constraint ((Owner=?="aeolus"&&JobUniverse==9)) && ((Managed =!= "ScheddDone")) && (JobStatus == 3 || JobStatus == 4 || (JobStatus == 5 && Managed =?= "External")) 05/23/11 12:30:49 [18785] Fetched 0 job ads from schedd 05/23/11 12:30:49 [18785] Updating classad values for 6.0: 05/23/11 12:30:49 [18785] EnteredCurrentStatus = 1306150249 05/23/11 12:30:49 [18785] HoldReason = "Create_Instance_Failure: 500 Internal Server Error" 05/23/11 12:30:49 [18785] HoldReasonCode = 0 05/23/11 12:30:49 [18785] HoldReasonSubCode = 0 05/23/11 12:30:49 [18785] JobStatus = 5 05/23/11 12:30:49 [18785] Managed = "Schedd" 05/23/11 12:30:49 [18785] NumSystemHolds = 1 05/23/11 12:30:49 [18785] ReleaseReason = undefined 05/23/11 12:30:49 [18785] No jobs left, shutting down 05/23/11 12:30:49 [18785] leaving doContactSchedd() 05/23/11 12:30:49 [18785] Got SIGTERM. Performing graceful shutdown. 05/23/11 12:30:49 [18785] Started timer to call main_shutdown_fast in 1800 seconds 05/23/11 12:30:49 [18785] **** condor_gridmanager (condor_GRIDMANAGER) pid 18785 EXITING WITH STATUS
It is hard to say if this is the exact problem.
The problem I still know about is that condor generates instance names like:
Condor_localhost.localdomain_job#12.0 (or something like that)
Unfortunately, RHEV-M VM names can only be 50 characters, and the above is always longer than 50 characters. Whether that causes the above error is unclear to me, but until we fix that, RHEV-M is not going to work.