Storage:
FreeNAS-8.3.1-Release-X64
Supermicro X9DRI-LN4F Intel E5-2620
DDR3-1600 ECC LSI 9211-8i 32GB
Intel 520 120G 7.2K Drives*17
Intel Pro 1000MT Quad Port
LAN: 6 port with mode 0 bonding
Servers:
RHEL 5.5 x64
LAN: 2 port with mode 0 bonding
switch:
H3C 5700 Gigabyte switch
Storage pool
16 2T HDD RAID10 with 16TB size, zfs, compress=lzjb, atime=off
snapshot every day and keep 2weeks, every weekend keep 2 months, total about 600
snapshot.
Total is about 10 million files, 200 thousands folders, 80% data is about 1~4K small files, ave size is about 8K. Total size is about 10TB. 60% write & 40% read.
Every user have their datasheet, and mount on every servers separate.
The read and write is seem OK, IOPS can top to 5K, but there have lockd problem several times. The client have no respond when access their folder, and there has a large "Server Ret-Failed" when show "nfsstat -s" on storage, and error "kernel: lockd: server not responding, timed out " on servers side.
After increate "numbers of servers" , it can clam the no respond problem, now the numbers of servers is 200.
If any way to resolve such problem?
If there any problem when close lockd with -o nolock? I have set this on one server, but some application need lock function.
If any advice to improve performance with such surrounding more. We need more 30% performance last half year.
Thanks a lot.
BTW, I have another pool hardware backup storage , it work well with zfs and gzip. It can do full backup in 24Hour when old backup need 3 days when use ext4.
FreeNAS-8.3.1-Release-X64
Supermicro X9DRI-LN4F Intel E5-2620
DDR3-1600 ECC LSI 9211-8i 32GB
Intel 520 120G 7.2K Drives*17
Intel Pro 1000MT Quad Port
LAN: 6 port with mode 0 bonding
Servers:
RHEL 5.5 x64
LAN: 2 port with mode 0 bonding
switch:
H3C 5700 Gigabyte switch
Storage pool
16 2T HDD RAID10 with 16TB size, zfs, compress=lzjb, atime=off
snapshot every day and keep 2weeks, every weekend keep 2 months, total about 600
snapshot.
Total is about 10 million files, 200 thousands folders, 80% data is about 1~4K small files, ave size is about 8K. Total size is about 10TB. 60% write & 40% read.
Every user have their datasheet, and mount on every servers separate.
The read and write is seem OK, IOPS can top to 5K, but there have lockd problem several times. The client have no respond when access their folder, and there has a large "Server Ret-Failed" when show "nfsstat -s" on storage, and error "kernel: lockd: server not responding, timed out " on servers side.
After increate "numbers of servers" , it can clam the no respond problem, now the numbers of servers is 200.
If any way to resolve such problem?
If there any problem when close lockd with -o nolock? I have set this on one server, but some application need lock function.
If any advice to improve performance with such surrounding more. We need more 30% performance last half year.
Thanks a lot.
BTW, I have another pool hardware backup storage , it work well with zfs and gzip. It can do full backup in 24Hour when old backup need 3 days when use ext4.