iSCSI - block storage over IP · Linux

iSCSI transports SCSI commands over TCP/IP, providing block storage over the network. It is the most common SAN protocol for Linux deployments, especially with cost-effective Ethernet instead of Fibre Channel.

How iSCSI works

iSCSI has two components: the target (the storage) and the initiator (the host). The initiator sends SCSI commands over TCP to the target, which executes them on its disks and returns the results. The host sees the result as a local block device (/dev/sdX or /dev/mapper/mpathX).

Read-only / Safeiscsiadm session— iscsiadm -m session shows the active iSCSI sessions. Each session is one connection to a target. The iqn (iSCSI qualified name) identifies the target. Multiple sessions to the same target with different paths enable multipath.

$ iscsiadm -m session

tcp: [1] 10.0.0.10:3260,1 iqn.2000-01.com.synology:target-1 (non-flash)

Connecting to an iSCSI target

Configuration changeiscsiadm discovery and login— iscsiadm -m discovery -t sendtargets -p IP discovers targets at the portal. iscsiadm -m node -L all logs in to all discovered targets. After login, the new LUN appears as a block device. Use iscsiadm -m session to verify the active sessions.

$ iscsiadm -m discovery -t sendtargets -p 10.0.0.10; iscsiadm -m node -L all

10.0.0.10:3260,1 iqn.2000-01.com.synology:target-1
Logging in to [iface: default, target: iqn.2000-01.com.synology:target-1, portal: 10.0.0.10,3260] (multiple)
Login to [iface: default, target: iqn.2000-01.com.synology:target-1, portal: 10.0.0.10,3260] successful.

Mounting and using the LUN

Data-loss riskuse LUN— After iSCSI login, the new LUN appears as a block device (e.g. /dev/sdb or /dev/sdc). Use fdisk or parted to partition, mkfs to create a filesystem, mount to attach. The same workflow as a local disk.

$ lsblk; fdisk /dev/sdb; mkfs.ext4 /dev/sdb1; mount /dev/sdb1 /mnt/iscsi

Multipath for redundancy

Read-only / Safemultipath -ll— multipath -ll shows the multipath topology. The mpath0 device is composed of two paths (sda and sdb), each with two underlying targets (sdc, sdd). If one path fails, the I/O continues on the other. multipathd runs as a daemon and handles path failures automatically.

$ multipath -ll

mpath0 (3600508...) dm-3 SYNOLOGY,iSCSI Disk
size=1.0T features='1 queue_if_no_path' wp=rw
|-+- policy='round-robin 0' prio=1 status=active
| |- 1:0:0:0 sda 8:0   active ready running
| `- 2:0:0:0 sdb 8:16  active ready running
`-+- policy='round-robin 0' prio=1 status=enabled
|- 1:0:1:0 sdc 8:32  active ready running
`- 2:0:1:0 sdd 8:48  active ready running