0

When large read and write operations are conducted on a NFSv4 share on a CentOS 7.3 client, I notice that the NFS mounts often time out.

Does a fresh, out-of-the-box installation of CentOS 7.3 require letting the NFS service through the default firewall? If so, how?

Improve this question
3
  • Couldn't be related to different network settings on both ends? I think specifically about jumbo frames enabled on client side but not on server.
    – TerDale
    Sep 25, 2018 at 16:57
  • Jumbo frames are enabled on server, client, and switch. Something else wacky is going on.
    – user260467
    Sep 25, 2018 at 17:00
  • Which firewall?
    – Michael Hampton
    Sep 25, 2018 at 17:06

3 Answers 3

Reset to default
1

Theoretically, any network device like router or firewall may drop packages under load. However, it's more probable that your NFS server unable to handle large IO or your client is not optimally configured.

One of the problems that we have observed, that if client has lot of memory, then flush of dirty pages happens to late. As a result we had lot of nfs timeouts. Try to set

vm.dirty_background_bytes
vm.dirty_bytes

to reasonable numbers, like 500mb and 1GB. The defaults are ~ 10% and 20%, which might be too much.

Improve this answer
0

Either the firewall lets the traffic pass or it blocks it. It won’t randomly block it only sometimes. If you can mount the share and transfer data, you have some other issues.

Improve this answer
1
  • Unlikely in this case, but a stateful firewall will block traffic only sometimes. You can open a connection and transfer data and still have some of your packets blocked.
    – doneal24
    Sep 27, 2018 at 12:16
0

Turns out that this was a bug in the application itself on the client. Developer fixed it with a software update. NFS was configured properly on client, switch, and server.

Improve this answer

You must log in to answer this question.

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