On the workernode Things you can try first (assuming it's the LZ repo you are after):
cvmfs_talk -i lz.opensciencegrid.org host probe
next escalation step: cvmfs_config wipecache
In emergencies, try:
cvmfs_config killall
service autofs restart (if that does not work: "reboot")