We have a number of client machines connected to an NFS server. Right now, we mount the server's 10.0.0.xxx:/mnt/storage onto /mnt/fileserver on each client. The mounted storage contains a large number of subdirectories.
We're considering a change that would mean that we mount each subdirectory independently on the client end:
10.0.0.xxx:/mnt/storage/a->/mnt/fileserver/a10.0.0.xxx:/mnt/storage/b->/mnt/fileserver/b10.0.0.xxx:/mnt/storage/c->/mnt/fileserver/c10.0.0.xxx:/mnt/storage/d->/mnt/fileserver/d10.0.0.xxx:/mnt/storage/e->/mnt/fileserver/e
The plan is that /etc/exports on the server would still contain /mnt/storage — we would not change it to list all of the subdirectories.
There would be a large number of these separate mounts — probably 300 or so for each client machine.
Would this increase the number of client-server connections, and thus increase the load on the server? Or does NFS do some kind of aggregation of all mounts to the same export on the same server down to one underlying connection?
We're using nfs version 3, with clients being a mixture of Ubuntu 16.04 and 14.04, and the servers are 14.04.
Best Answer
It will not take that much load on the sever but the nfsd daemon will take time while starting and stopping.You can use automounter in that case.You can refer this automounter link