0

From the Linux kernel-parameters.txt I have seen that it is possible to change the NFS pool behaviour to have the nfsd threads bound to NUMA zones. The parameter in question is 

sunrpc.pool_mode

and can be set to pernode for NUMA affinity. There is also a note about NICs and their interrupts.

Depending on how many NICs you have and where their interrupts are bound, this option will affect which CPUs will do NFS serving.

I am not sure what how to understand this. Does it mean that if you e.g. have a 2-node NUMA system with 1 NIC, only the NFS threads that are bound to same the NUMA node as the NICs interrupts will be active and serving requests?

Hope someone already made some experience with this.

Thomas

Thomas
  • 4,155
  • 5
  • 21
  • 28
  • 1
    What's your goal? – ewwhite Feb 17 '19 at 17:01
  • @ewwhite: I was thinking of applying this setting as it seems to make sense on a NUMA system, but not sure about any implications. Maybe it is a bad idea at all? Unfortunately I do not have a test environment to test and this setting cannot be changed on the fly. – Thomas Feb 17 '19 at 17:55
  • 1
    I guess I mean to ask, “what problem are you looking to solve with this?” And if you have an application for this, why not test and observe the effect? – ewwhite Feb 17 '19 at 23:50
  • @ewwhite: agree. There is no real problem I try to solve, was just curious about this option and thought someone might have experiences with that already. – Thomas Feb 18 '19 at 07:31

0 Answers0