atlas-connect-l AT lists.bnl.gov

Subject: Atlas-connect-l mailing list

List archive

[Atlas-connect-l] AUTH failed for files on MWT2_UC_LOCALGROUPDISK

From: Giordon Stark <gstark AT cern.ch>
To: atlas-connect-l <atlas-connect-l AT lists.bnl.gov>
Cc: "support AT connect.usatlas.org" <support AT connect.usatlas.org>, Ian Snyder <ian.michael.snyder AT cern.ch>
Subject: [Atlas-connect-l] AUTH failed for files on MWT2_UC_LOCALGROUPDISK
Date: Wed, 8 Nov 2017 04:54:28 +0000

Hi all,

I'm seeing a few auth failures that are definitely reproducible when using condor. It's not clear to me why the condor nodes are seeing these failures but I'm unable to reproduce it on the interactive node. Here's a gist of some of the failure logs: https://gist.github.com/kratsg/b34aadfe28f40e559f1357ca0296e242

The sites all seem different, and the dataset holding these files is this one: user.gstark:user.gstark.147910.Pythia8_AU2CT10_jetjet_JZ0W.e2403_s3142_s3143_r9589.mu200.v30_compile_OUTPUT

Any ideas? The biggest issue is that using EventLoop + CondorDriver, we don't necessarily have DAGman so it is not very easy to retry individual jobs (especially when grouped in 10 files per job) and it would be great to reduce the AUTH errors here [which don't make sense anyway to me].

Giordon

[Atlas-connect-l] AUTH failed for files on MWT2_UC_LOCALGROUPDISK, Giordon Stark, 11/07/2017