Score:0

NFSv4 sharing directories over different NICs fails

iq flag

i am surprised to run into problems with the following attempt:

#/etc/exports
/path/to/dir1 10.10.2.3 (...)
/path/to/dir2 10.10.4.10 (...)
# server host
ip a
eth0 10.10.2.20
eth1 10.10.4.20
# client host
ip a
eth0 10.10.2.3
eth1 10.10.4.10

Setup:

  • Server and client are configured both for NFSv4.2 only.
  • Serverhost and client host both have a NIC in each of the above networks.
  • Background: i hoped to improve performance and reliability by isolating the traffic generated by access to one of both directories, where bandwith saturation may occur.

Problem:

Only one of each share can be mounted at a time (each of them work individually equally well), but not both of them at the same time, from the same client.

Any attempt to mount the second share always throws:

mount.nfs: mounting 10.10.{2|4}.20:/path/to/dir{1“} failed, \
reason given by server: No such file or directory

Is this supposed to fail?

I expected to be able to access the shares from same client host via different NICs.

Stkel avatar
iq flag
I just realised that it only fails when specifying "ver 4.2" to the mount command, while it actually works, if specifying "vers=4.0"! Therefore, the "trunking" feature of NFSv4 may be a reason?
Score:0
iq flag

one preliminary answer is: this is supposed to fail for protocol version 1 and 2, because they implement trunking (apparently activated by default), which depends on having different ip addresses on same exports.

Apparently the NFS server falls back to refusal of additional connections to additional IP addresses which are not referring to the same share.

I find documentation on this topic scarce, in particular for the man pages and believe it would help tremendously if that would be added, as i see tremendous potential of the trunking feature for specific use cases.

I sit in a Tesla and translated this thread with Ai:

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.