Resolve latch contention within the PFS

If the pfs_qualifier field displayed in the MOUNT LATCH ACTIVITY or FILE SYSTEM LATCH ACTIVITY table shows either Running or Osi Wait, do the following steps depending on the type of file system in question:

For a zFS file system, do the following (see z/OS Distributed File Service zFS Administration for more information):
  1. Issue the DISPLAY ZFS,QUERY,THREADS to get details about the state of threads within zFS.
  2. Issue modify zfs,hangbreak to have zFS to post any requests in zFS that are waiting, with a failure. This can allow the hang condition to be broken and resolved.
  3. Issue the modify zfs,query,threads command to determine if one or more requester threads remain in the same wait over several queries.
  4. If you cannot successfully break or resolve the hang, go to Terminate or cancel the latch holder
For an NFS client, most problems relate to the socket sessions that NFS has with its servers. Do the following:
  1. Issue the DISPLAY OMVS,FILE command to show the MOUNT PARM= value specified when the file system was mounted. This value contains the name of the remote server and the remote directory path name for the file system where the file is.
  2. Find out whether the delay is in the remove server or the NFS client. To find out, issue the DISPLAY TCPIP,,NETSTAT,CONN command to display the state of the socket sessions between NFS and the remote server (which is always port number 2049). If the output shows that there are no socket sessions between the NFS client and the remote server, either the remote system or the remote server might be down. However, note that NFS does not always have persistent socket sessions for its servers, so the absence of sessions may just mean that socket sessions are not needed at this exact time.

    If the remote system or server is down and the file system was hard mounted, NFS will try to establish contact indefinitely. In that case, you might have to unmount the file system to free up the users.

    You can also use the shell ping command to check for connectivity to the remote system. Use display commands on to the remote server's system to see why the server is not responding. If the remote system is also a z/OS® system, you can diagnose the latch contention on the remote system starting with step 1.

For the DFS Glue module, IOEGLUE, the wait usually means that the file system is exported by the SMB or DFS server and that the file being accessed is being shared with some remote client. Issue the DISPLAY OMVS,A=ALL command to display the SERVER= line that can help to find the address space and process id of the server involved. See z/OS Distributed File Service SMB Administration for diagnosing SMB problems like hangs.

If you can not free the latch holder with these methods, you might need to force the process to terminate, or cancel the latch holder. See Terminate or cancel the latch holder for more instructions.