Hello all : I have a opensolaris snv_40 x86 server acts a nfs server and it shares ZFS pool filesystem (total space about 6.8T ) to 20 gentoo Linux clients and 15 soalrs9 x86 clients. Our server farm has been using since 1year Gigabit switch to connect erver servers and they are a isoluted subnet.
Many times, the opensolaris NFS got hang then panic for clients daily backup thougth NFS in the midnight. And then I check /var/adm/message and it's showen some error as following : [b] Jun 2 03:58:15 zfs /usr/lib/nfs/nfsd[1446]: [ID 576340 daemon.error] Cannot establish transport over /dev/tcp Jun 2 03:58:15 zfs /usr/lib/nfs/nfsd[1446]: [ID 557088 daemon.error] t_open /dev/tcp failed: t_errno 8, Out of stream resou rces Jun 2 03:58:15 zfs /usr/lib/nfs/nfsd[1446]: [ID 576340 daemon.error] Cannot establish transport over /dev/tcp Jun 2 03:58:25 zfs /usr/lib/nfs/nfsd[1446]: [ID 557088 daemon.error] t_open /dev/tcp failed: t_errno 8, Out of stream resources [/b] I had modified /etc/system and /etc/default/nfs : ?/etc/system set ip:ipcl_conn_hash_size=262144 set ncsize=2097152 set nfs:nfs3_max_transfer_size=65536 set nfs:nfs3_bsize=65536 ?/etc/default/nfs NFSD_SERVER=2048 and After server boot, I manally execute ndd command to increase max tcp connections: [b] ndd -set /dev/tcp tcp_conn_req_max_q 2048 ndd -set /dev/tcp tcp_conn_req_max_q0 2048 [/b] Sometime, I execute some of commands in the opensolaris NFS server when clients are backing up I got a strange error : [b] -bash: fork: Resource temporarily unavailable [/b] I think the problem that may opensolaris NFS server lacks of system resource. But I have no idea to tune the OS kernel parameter except /etc/system that showed as upon. set ip:ipcl_conn_hash_size=262144 set ncsize=2097152 set nfs:nfs3_max_transfer_size=65536 set nfs:nfs3_bsize=65536 Does anyone know how to tune OS kernel and other parameter that can improve NFS server reliability and performance??? This message posted from opensolaris.org
