Created attachment 589230 [details] Glusterfs logs Description of problem: Customer is getting an error when trying to NFS mount a sub-directory from a Solaris client. The glusterfs logs are attached. There was a suggestion to use "mountproto=tcp" option but that is not an available option in Solaris. gluster> volume info Volume Name: tvol1 Type: Replicate Volume ID: 6faf2030-fea4-4398-a039-1d29999d8bae Status: Started Number of Bricks: 1 x 2 = 2 Transport-type: tcp Bricks: Brick1: 172.17.9.231:/brick1 Brick2: 172.17.9.232:/brick1 Options Reconfigured: nfs.register-with-portmap: on nfs.rpc-auth-allow: 10.125.74.90 nfs.addr-namelookup: off nfs.disable: off nfs.export-dir: /gino features.limit-usage: /gino:10gb features.quota: on gluster> root@nysolsbx1:~# mount -o proto=tcp,vers=3 nfs://10.125.74.106:38467/tvol1/gino /mnt nfs mount: 10.125.74.106: : RPC: Program not registered nfs mount: retrying: /mnt Version-Release number of selected component (if applicable): RHS 2.0 Beta2. How reproducible: Always.
For now, we only have a workaround: Hi Rajesh, Thank you for the workaround, it works fine. Can we expect a patch in the future release to avoid this procedure ? Regards, Anthony > Date: Thu, 19 Jul 2012 05:30:24 -0400 > From: rajesh > To: sokar6012 > CC: gluster-users; sgowda > Subject: Re: [Gluster-users] NFS subdirectory Solaris client > > fwding from krishna's mail, need to update the docs with this info: > > Here are the steps: > on linux client machine: > * make sure you have the directory you want to export is already created, if > not create it. > if /mnt/nfs is the mount point do this: > mkdir /mnt/nfs/subdir > * on storage node do: > gluster volume set <volname> nfs.export-dir /subdir > gluster volume set <volname> nfs.mount-udp on > * do a "showmount -e <storage-node-ip>" to see that subdir is exported too. > * on a LINUX client do this: > mount -o proto=tcp <storage-node-ip<volname>/subdir /mnt/nfs > i.e you are mounting the subdir. > * now on SOLARIS client do this: > mount nfs://<storage-node-ip<volname>/subdir /mnt/nfs > you should be able to access the exported subdir on Solaris machine. note > that you have to mount the subdir on a linux machine first with proto=tcp > before trying to mount on solaris machine.
For now we have the workaround mentioned above. We won't need the workaround when we implement the non blocking UDP-layer. Right now we have only TCP implementation. Solaris nfs client uses UDP for MOUNT operations.
*** Bug 858414 has been marked as a duplicate of this bug. ***
Based on discussion with Sayan, concluded this is not a supported OS.