NFS data corruption while scp a file on a remote host nfs mounted filesystem
Issue
The setting involve 3 machines:
-
metro a 'undefined' machine
-
qualif, a Red Hat Enterprise Linux 5 box.
-
filer, a nfs filer.
-
qualif is mounting a share exported by filer:
filer:/vol/QAKRM01 on /data/ora/PKRM01 type nfs (rw,nosuid,nodev,nfsvers=3,tcp,bg,hard,nointr,nolock,rsize=32768,wsize=32768,timeo=600)
- A file has been created randomly with on metro with :
dd if=/dev/urandom of=test.dump bs=1024 count=10000000
-
If user foo, logged on metro, scp test.dump on qualif in his home directory, the file will be there with the same md5sum as the one computed on metro.
-
If user foo logged on metro, scp test.dump on qualif in /data/ora/PKRM01 (the share physically hosted on filer), then the md5sum of the resulting file will not match the original file.
Environment
- Red Hat Enterprise Linux 5
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.