[CentOS] automounted filesystem appears empty on first reference

Dan Halbert

halbert at everyzing.com
Thu Oct 4 03:33:47 UTC 2007


I have an odd automount problem. The first time I look in an automounted 
filesystem (when the mount is requested), it is empty; the second time 
it is not. Moreover, I only see this behavior on Centos 4.x systems, but 
not on Centos 5.0 systems going to the same fileserver. Here's an 
example, with some details simplified to clarify. In this example, 
fileserver1 is exporting fs1.

    centos4$ /bin/ls /mnt/fileserver1/fs1      # first time, directory 
looks empty
    centos4$ /bin/ls /mnt/fileserver1/fs1      # second time, I see the 
contents
    file1   dir2

I don't see this on Centos 5, which comes with a new (v5.x vs v4.x) 
version of automount:

    centos5$ /bin/ls /mnt/fileserver1/fs1      # I see the contents 
right away, on the first try
    file1   dir2

I watched the logs on fileserver1, and automount request is logged 
immediately after the first ls prints nothing.

The automount maps are stored in ldap. The Centos4 and Centos5 systems 
are referencing the same maps. The entry in auto.master for fileserver1 
(for cn=/mnt/fileserver1) is:

    ldap:ou=auto.fileserver1,ou=autofs,dc=example,dc=com --timeout=86400 
--ghost -o 
rw,hard,async,noatime,intr,retrans=4,timeo=100,rsize=8192,wsize=8192

and in auto.fileserver1 is (for cn=*):

    fileserver1.example.com:/export/&


At first I thought it had to do with the --ghost, but I took that out 
and adjusted things and still had the same problem. It is completely 
reproducible if I umount the automount between tries.

Any idea what is causing this? I have been unable to find a 
similar-sounding problem, even with extensive Googling. Why does the 
first automount appear to be asynchronous, with some kind of race 
condition, and the second is synchronous? I even see a little hesitation 
on the centos5 box before ls prints the directory contents, so it 
appears to be waiting for the automount to complete. All the Centos 4 
and Centos 5 boxes are current with "yum update".

This problem is causing failures on some of our applications because 
they don't see their data the first time they run, when the automount occurs

We see this across many boxes, and several fileservers (the fileservers 
are Fedora Core 5 and 6 machines, for stupid reasons).

Thanks,
Dan

   







More information about the CentOS mailing list