There is something I need to understand about NFS.
If I dd (dd if=“X” of=*Y") a large partition and write to NFS on another server the image is corrupt. If I just dd the same partition to a local drive on the machine, then it is not corrupt. I never had this problem during the previous decade!!
Clearly NFS is to blame. Is there any way that I maybe mounted my NFS partitions in such a way that error correction was not enabled ?
NFS is not meant for very large files. It is perfect for small files sharing. NFS does lots of things behind such as caching and other stuff hence you are getting such weird results.
Thats weird. I have been doing the exact thing for 10++ years and never had these problems. It is clearly a new NFS thing. I could always write to an NFS volume with dd without any loss. Getting any help from the NFS developers is moot. They never answer, so I will have to try and figure this out myself somehow.
In that case, it is seriously broken. Unreported Data Corruption is THE cardinal sin of a file system. A file system MUST either complete the transaction faithfully, or abort with an error code/message.
I downgraded a few versions and the problem went away. I cannot try the newest version as there will be dependency problems, but it seems that for some reason the NFS versions and particularly updates that came with my distro I use on these servers did not make hard mount but did soft mounts in place of hard mounts automatically. It is the only explanation that makes sense. So it seems the “-s” switch was default even if you do not specify it.
It works perfect now again.
Unfortunately there is about no way to verify whether you have a hard linked NFS or soft linked NFS volume I know of. that would be helpful to add to cron to report oif a soft linked nfs volume exists, but I have no clue how to test for that at the moment.