iscsi、nfs与ceph
一、iscsi设置
1. 服务器端
| [1] | Install administration tools. |
[2] Configure iSCSI Target .
For example, create a disk image under the [/iscsi_disks] directory and set it as a shared disk.
# create a disk image
[root@dlp ~]# mkdir /iscsi_disks
[root@dlp ~]# dd if=/dev/zero of=/iscsi_disks/disk01.img count=0 bs=1 seek=80G
[root@dlp ~]# vi /etc/tgt/targets.conf
# add follows to the end
# if you set some devices, add <target>-</target> and set the same way with follows
# naming rule : [ iqn.yaer-month.domain:any name ]
<target iqn.2014-08.world.server:target00>
# provided devicce as a iSCSI target
backing-store /iscsi_disks/disk01.img
# iSCSI Initiator's IP address you allow to connect
initiator-address 10.0.0.31
# authentication info ( set anyone you like for "username", "password" )
incominguser username password
</target>
[root@dlp ~]# /etc/rc.d/init.d/tgtd start
Starting SCSI target daemon: [ OK ]
[root@dlp ~]# chkconfig tgtd on
# confirm status
[root@dlp ~]# tgtadm --mode target --op show
Target 1: iqn.2014-08.world.server:target00
System information:
Driver: iscsi
State: ready
I_T nexus information:
LUN information:
LUN: 0
Type: controller
SCSI ID: IET 00010000
SCSI SN: beaf10
Size: 0 MB, Block size: 1
Online: Yes
Removable media: No
Prevent removal: No
Readonly: No
Backing store type: null
Backing store path: None
Backing store flags:
LUN: 1
Type: disk
SCSI ID: IET 00010001
SCSI SN: beaf11
Size: 85899 MB, Block size: 512
Online: Yes
Removable media: No
Prevent removal: No
Readonly: No
Backing store type: rdwr
Backing store path: /iscsi_disks/disk01.img
Backing store flags:
Account information:
username
ACL information:
10.0.0.31 2. 客户端配置
[1] Configure iSCSI Initiator.
[root@www ~]# yum -y install iscsi-initiator-utils [root@www ~]# vi /etc/iscsi/iscsid.conf # line 53: uncomment node.session.auth.authmethod = CHAP # line 57,58: uncomment and set username and password which set on iSCSI Target node.session.auth.username = username node.session.auth.password = password # discover target [root@www ~]# iscsiadm -m discovery -t sendtargets -p 10.0.0.30 Starting iscsid: [ OK ] 10.0.0.30:3260,1 iqn.2014-08.world.server:target00 # confirm status after discovery [root@www ~]# iscsiadm -m node -o show # BEGIN RECORD 6.2.0-873.10.el6 node.name = iqn.2014-08.world.server:target00 node.tpgt = 1 node.startup = automatic node.leading_login = No ... ... ... node.conn[0].iscsi.IFMarker = No node.conn[0].iscsi.OFMarker = No # END RECORD # login to target [root@www ~]# iscsiadm -m node --login Logging in to [iface: default, target: iqn.2014-08.world.server:target00, portal: 10.0.0.30,3260] (multiple) Login to [iface: default, target: iqn.2014-08.world.server:target00, portal: 10.0.0.30,3260] successful. # confirm established session [root@www ~]# iscsiadm -m session -o show tcp: [1] 10.0.0.30:3260,1 iqn.2014-08.world.server:target00 # confirm partitions [root@www ~]# cat /proc/partitions major minor #blocks name 8 0 209715200 sda 8 1 512000 sda1 8 2 209202176 sda2 253 0 200966144 dm-0 253 1 8232960 dm-1 8 16 83886080 sdb # added new device provided from target as [sdb][2] It's possible to use iSCSI device like follows.
[root@www ~]# yum -y install parted
# create a label
[root@www ~]# parted --script /dev/sdb "mklabel msdos"
# create a partition
[root@www ~]# parted --script /dev/sdb "mkpart primary 0% 100%"
# format with EXT4
[root@www ~]# mkfs.ext4 /dev/sdb1
# mount
[root@www ~]# mount /dev/sdb1 /mnt
[root@www ~]# df -hT
Filesystem Type Size Used Avail Use% Mounted on
/dev/mapper/vg_dlp-lv_root
ext4 189G 1.1G 179G 1% /
tmpfs tmpfs 1.9G 0 1.9G 0% /dev/shm
/dev/sda1 ext4 485M 75M 385M 17% /boot
/dev/sdb1 ext4 79G 184M 75G 1% /mnt
补充：
客户端安装后需要启动守护进程，然后进行发现：
/etc/init.d/iscsid start iscsiadm -m discovery -t st -p 12.123.0.51 iscsiadm -m discovery -T iqn.2015-06.com.oracle:zjxl -p 12.123.0.51:3260 -l #修改配置文件 vim /var/lib/iscsi/send_targets/12.123.0.51\,3260/iqn.2015-06.com.oracle\:oracle\:zjxl\,12.123.0.51\,3260\,1\,default/default配置multipath：
一个机器至少要挂2个存储IP，一个坏了，另一个还可以连接。编辑/etc/multipath.conf
blacklist{
devnode "^sda"
}
defaults{
user_friendly_names yes
udev_dir /dev
path_grouping_policy multibus
failback immediate
no_path_retry fail
} blacklist代表系统盘不做multipath。
然后service multipathd restart；执行cat /proc/partitions可以看到以dm-*开头的设备，这些就是多路径设备了。
二、ceph设置
其实对于ceph来讲，直接通过rbd create一个块，然后rbd map这个块，再rbd showmapped可以看到此块map的路径，如/dev/rbd0。然后在iscsi服务器端的/etc/tgt/targets.conf进行配置：
<target iqn.2008-09.com.example:server.target11>
direct-store /dev/rbd0
</target> 这样在iscsi客户端就能使用此块设备了。
但这样使用的代价比较大，因为它是通过ceph rbd kernel module的形式，挂载rbd。这样会有大量的内核态和用户态的切换，势必会影响性能。
这个问题已经有人解决了：
http://ceph.com/dev-notes/adding-support-for-rbd-to-stgt/
http://ceph.com/dev-notes/updates-to-ceph-tgt-iscsi-support/
http://www.sebastien-han.fr/blog/2014/07/07/start-with-the-rbd-support-for-tgt/
在ubuntu上的iscsi已经支持rbd；fedora 20之后有支持rbd的iscsi rpm包。但在centos 6.5，无法配置带rbd支持的iscsi。
如果你也配置过此功能，欢迎交流。
三、利用支持rbd的scsi-target-utils
利用上述方法使ceph导出iscsi设备是没有问题的，但由于其利用了rbd的内核模块，导致内核态和用户态的频繁切换，势必会影响性能。可不可以使rbd块设备直接被导出成为iscsi设备呢？答案是肯定的。
先从以下地址下载支持rbd的scsi-target-utilsrpm包：
#wget http://ceph.com/packages/ceph-extras/rpm/centos6/x86_64/scsi-target-utils-1.0.38-48.bf6981.ceph.el6.x86_64.rpm #rpm -ivh scsi-target-utils-1.0.38-48.bf6981.ceph.el6.x86_64.rpm安装完成后，查看当前tgt对于rbd driver是否支持：
# tgtadm --lld iscsi --mode system --op show
System:
State: ready
debug: off
LLDs:
iser: error
iscsi: ready
Backing stores:
rbd (bsoflags sync:direct)
rdwr (bsoflags sync:direct)
ssc
null
bsg
sg
sheepdog
Device types:
passthrough
tape
changer
controller
osd
cd/dvd
disk
iSNS:
iSNS=Off
iSNSServerIP=
iSNSServerPort=3205
iSNSAccessControl=Off 创建一个rbd设备：
#rbd create iscsi/tgt1 -s 10240修改/etc/tgt/targets.conf，导出刚才创建的rbd设备：
include /etc/tgt/conf.d/*.conf <target iqn.2014-11.rbdstore.com:iscsi> driver iscsi bs-type rbd backing-store iscsi/tgt1 </target>重启tgt：
# /etc/init.d/tgtd restart Stopping target framework daemon Starting target framework daemon
在iscsi initiator端连接该iscsi target：
[root@ceph-osd-1 ~]# iscsiadm -m discovery -t sendtargets -p 10.10.200.165
Starting iscsid: [ OK ]
10.10.200.165:3260,1 iqn.2014-11.rbdstore.com:iscsi
[root@ceph-osd-1 ~]# iscsiadm -m node -T iqn.2014-11.rbdstore.com:iscsi -l
Logging in to [iface: default, target: iqn.2014-11.rbdstore.com:iscsi, portal: 10.10.200.165,3260] (multiple)
Login to [iface: default, target: iqn.2014-11.rbdstore.com:iscsi, portal: 10.10.200.165,3260] successful.
[root@ceph-osd-1 ~]# fdisk -l
Disk /dev/sdb: 5788.2 GB, 5788206759936 bytes
255 heads, 63 sectors/track, 703709 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00000000
Disk /dev/sda: 209.7 GB, 209715068928 bytes
255 heads, 63 sectors/track, 25496 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x0009a9dd
Device Boot Start End Blocks Id System
/dev/sda1 * 1 131 1048576 83 Linux
Partition 1 does not end on cylinder boundary.
/dev/sda2 131 1176 8392704 82 Linux swap / Solaris
/dev/sda3 1176 25497 195356672 8e Linux LVM
Disk /dev/mapper/vg_swift-LogVol00: 200.0 GB, 200043134976 bytes
255 heads, 63 sectors/track, 24320 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00000000
Disk /dev/mapper/docker-253:0-3539142-pool: 107.4 GB, 107374182400 bytes
255 heads, 63 sectors/track, 13054 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 65536 bytes
Disk identifier: 0x00000000
Disk /dev/sdc: 10.7 GB, 10737418240 bytes
64 heads, 32 sectors/track, 10240 cylinders
Units = cylinders of 2048 * 512 = 1048576 bytes
Sector size (logical/physical): 512 bytes / 4194304 bytes
I/O size (minimum/optimal): 4194304 bytes / 4194304 bytes
Disk identifier: 0x00000000
从以上可以看到，导出成功。
四、nfs设置
在ceph的一个节点利用rbd map一个块设备，然后格式化并挂载在一个目录，如/mnt。在此节点上安装nfs的rpm包：
yum -y install nfs-utils设置挂载目录：
[root@mon0 mnt]# cat /etc/exports /mnt 192.168.101.157(rw,async,no_subtree_check,no_root_squash) /mnt 192.168.108.4(rw,async,no_subtree_check,no_root_squash)启动并导出：
service rpcbind start chkconfig rpcbind on service nfs start chkconfig nfs on exportfs -r客户端查看一下：
[root@osd2 /]# showmount -e mon0 Export list for mon0: /mnt 192.168.108.4,192.168.101.157然后挂载：
mount -t nfs mon0:/mnt /mnt
需要注意的是，NFS默认是用UDP协议，如果网络不稳定，换成TCP协议即可：
mount -t nfs mon0:/mnt /mnt -o proto=tcp -o nolock
五、rbd-fuse设置
在客户机上配置ceph.repo后安装rbd-fuse的rpm包，然后就可以挂载pool了：
rbd-fuse -p test /mnt上面的示例是在没有cephx下将test pool挂载到客户机的/mnt。然后就可以在/mnt中看到test pool中的块了。此时可以利用losetup挂载这个img。
卸载直接利用fusermount -u /mnt。
| |