IssueSrmWrongTurl

Symptoms

Occurrences

Observations

SRM read requests from external sources keep hanging and timeout. Looking at the debug output reveals that the returned TURL contains a gridFTP door host name without domain extension:

srmcp -2 -debug srm://t3se01.psi.ch:8443/srm/managerv2?SFN=//pnfs/psi.ch/cms/testing/derek file:////tmp/derek501 | grep "gsiftp://"
Storage Resource Manager (SRM) Client version 2.0.9
Copyright (c) 2002-2008 Fermi National Accelerator Laboratory

copying CopyJob, source = gsiftp://t3fs01:2811//pnfs/psi.ch/cms/testing/derek destination = file:////tmp/derek501

This happens consistently for certain fileservers, but not for all of them.

The /etc/hosts file on an affected fileserver has an entry with a local name, only. dCache regrettably only does a local name resolution:

192.33.123.41  t3fs01  # Added by DHCP

Solution or Workaround

Either one needs to comment the /etc/hosts entry out or modify it by adding a leading FQDN. Need to study how to convince DHCP to make the right entry automatically.

Another possibility is to use a static network setup for the File servers. Probably anyhow a good idea.

Monitoring for this condition

-- DerekFeichtinger - 11 May 2009