I have a java app that accesses an NFS share with hard mount (soft mount does troubles i get many RPC timeout's for some reason). I'm on solaris. If i turn off (svcadm -v disable to the nfs server) then my java app gets stuck if doing
kill -3 myPID
returns nothing as my java process hangs - strange, why even kill -3 to get thread dump does not return? also i even had problems killing my process - its not something i want to do, i don't want my app to get stuck if the mount is down Anyone has a solution or can recommend what I should be doing? I would like to use maybe soft mounts the problem is that i get many RPC timeout while I try to copy files from one server to an nfs share.
NFS write failed for server xx.xx.xx.xx: error 5 (RPC: Timed out)
cp: /share/myfile.txt: Connection timed out
so because i'm getting these errors with soft mounts and it looks unreliable (its over tcp) then i wished to test hard mount but this makes my app stuck if the share is down for some reason.
Blocking-and-retrying-indefinitely is the very definition of a hard-mount. Passing the
introption to the mount command ought to at least allow you to interrupt the blockage when the mount is down, if I'm reading the docs correctly.More details here.
Two problems there. For your app, get a stack trace from your threads and see where they are hanging. Perhaps they are indirectly touching something that depends on a NFS path (eg. is any NFS-mounted directories in your PATH variable?).
The RPC timeout could be caused by network instability, or your client overloading the server, etc. You'll have to investigate it carefully. Soft mounts over TCP work just fine and we use it extensively.