2

I have a bunch of compute clients working on jobs, saving output data to a NAS machine. All machines are centos 6.2. They mount it via automount NFS, with a timeout of 1200 (default config).

The NAS machine needs to be restarted.

If I can restart the machine within that 1200s (20 minute) window, will the clients just block on IO until it comes back up? A minor interruption (pause) in service is ok, as long as it doesn't cause the running processes to error out.

If necessary I could loop through and SIGSTOP all job processes, restart and resume them -- I just don't want to break the open file handles.

How can I run a restart like this without killing processes with open files?

2
  • What NFS version are you using?
    – sciurus
    Oct 30, 2013 at 0:28
  • if you mount with option 'hard' , which is default, then your clients will block as long as nfs server not available. The automount timeout is option tells you how long is the idle time until nfs server unmounted.
    – kofemann
    Oct 30, 2013 at 9:06

1 Answer 1

1

They will just block I/O yes until available again, should be no need for SIGSTOP as NFS is fairly mature this was accounted for quite some years ago.

biod uses read ahead and write behind so your file handles should be unscathed by a reboot of the NFS server during the timeout if the timeout was even triggered to begin with, so I suppose what I am saying is that so many events would have to align to begin with for this to even become a question and even then the answer is no it will be perfectly fine :)

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .