iSCSI Target

iSCSI Target Getting Started Guide

The Storage Performance Development Kit iSCSI target application is named iscsi_tgt. This following section describes how to run iscsi from your cloned package.


This guide starts by assuming that you can already build the standard SPDK distribution on your platform.

Once built, the binary will be in app/iscsi_tgt.

If you want to kill the application by using signal, make sure use the SIGTERM, then the application will release all the shared memory resource before exit, the SIGKILL will make the shared memory resource have no chance to be released by applications, you may need to release the resource manually.


The following diagram shows relations between different parts of iSCSI structure described in this document.

iSCSI structure

Configuring iSCSI Target via config file

A iscsi_tgt specific configuration file is used to configure the iSCSI target. A fully documented example configuration file is located at etc/spdk/

The configuration file is used to configure the SPDK iSCSI target. This file defines the following: TCP ports to use as iSCSI portals; general iSCSI parameters; initiator names and addresses to allow access to iSCSI target nodes; number and types of storage backends to export over iSCSI LUNs; iSCSI target node mappings between portal groups, initiator groups, and LUNs.

You should make a copy of the example configuration file, modify it to suit your environment, and then run the iscsi_tgt application and pass it the configuration file using the -c option. Right now, the target requires elevated privileges (root) to run.

app/iscsi_tgt/iscsi_tgt -c /path/to/iscsi.conf

Assigning CPU Cores to the iSCSI Target

SPDK uses the DPDK Environment Abstraction Layer to gain access to hardware resources such as huge memory pages and CPU core(s). DPDK EAL provides functions to assign threads to specific cores. To ensure the SPDK iSCSI target has the best performance, place the NICs and the NVMe devices on the same NUMA node and configure the target to run on CPU cores associated with that node. The following command line option is used to configure the SPDK iSCSI target:

-m 0xF000000

This is a hexadecimal bit mask of the CPU cores where the iSCSI target will start polling threads. In this example, CPU cores 24, 25, 26 and 27 would be used.

Configuring a LUN in the iSCSI Target

Each LUN in an iSCSI target node is associated with an SPDK block device. See Block Device User Guide for details on configuring SPDK block devices. The block device to LUN mappings are specified in the configuration file as:

LUN0 Malloc0
LUN1 Nvme0n1

This exports a malloc'd target. The disk is a RAM disk that is a chunk of memory allocated by iscsi in user space. It will use offload engine to do the copy job instead of memcpy if the system has enough DMA channels.

Configuring iSCSI Target via RPC method

In addition to the configuration file, the iSCSI target may also be configured via JSON-RPC calls. See JSON-RPC Methods for details.

Portal groups

  • add_portal_group – Add a portal group.
  • delete_portal_group – Delete an existing portal group.
  • add_pg_ig_maps – Add initiator group to portal group mappings to an existing iSCSI target node.
  • delete_pg_ig_maps – Delete initiator group to portal group mappings from an existing iSCSI target node.
  • get_portal_groups – Show information about all available portal groups.
python /path/to/spdk/scripts/ add_portal_group 1

Initiator groups

  • add_initiator_group – Add an initiator group.
  • delete_initiator_group – Delete an existing initiator group.
  • add_initiators_to_initiator_group – Add initiators to an existing initiator group.
  • get_initiator_groups – Show information about all available initiator groups.
python /path/to/spdk/scripts/ add_initiator_group 2 ANY

Target nodes

  • construct_target_node – Add a iSCSI target node.
  • delete_target_node – Delete a iSCSI target node.
  • target_node_add_lun – Add an LUN to an existing iSCSI target node.
  • get_target_nodes – Show information about all available iSCSI target nodes.
python /path/to/spdk/scripts/ construct_target_node Target3 Target3_alias MyBdev:0 1:2 64 -d

Configuring iSCSI Initiator

The Linux initiator is open-iscsi.

Installing open-iscsi package Fedora:

yum install -y iscsi-initiator-utils


apt-get install -y open-iscsi


Edit /etc/iscsi/iscsid.conf

node.session.cmds_max = 4096
node.session.queue_depth = 128

iscsid must be restarted or receive SIGHUP for changes to take effect. To send SIGHUP, run:

killall -HUP iscsid

Recommended changes to /etc/sysctl.conf

net.ipv4.tcp_timestamps = 1
net.ipv4.tcp_sack = 0
net.ipv4.tcp_rmem = 10000000 10000000 10000000
net.ipv4.tcp_wmem = 10000000 10000000 10000000
net.ipv4.tcp_mem = 10000000 10000000 10000000
net.core.rmem_default = 524287
net.core.wmem_default = 524287
net.core.rmem_max = 524287
net.core.wmem_max = 524287
net.core.optmem_max = 524287
net.core.netdev_max_backlog = 300000


Assume target is at

iscsiadm -m discovery -t sendtargets -p

Connect to target

iscsiadm -m node --login

At this point the iSCSI target should show up as SCSI disks. Check dmesg to see what they came up as.

Disconnect from target

iscsiadm -m node --logout

Deleting target node cache

iscsiadm -m node -o delete

This will cause the initiator to forget all previously discovered iSCSI target nodes.

Finding /dev/sdX nodes for iSCSI LUNs

iscsiadm -m session -P 3 | grep "Attached scsi disk" | awk '{print $4}'

This will show the /dev node name for each SCSI LUN in all logged in iSCSI sessions.


After the targets are connected, they can be tuned. For example if /dev/sdc is an iSCSI disk then the following can be done: Set noop to scheduler

echo noop > /sys/block/sdc/queue/scheduler

Disable merging/coalescing (can be useful for precise workload measurements)

echo "2" > /sys/block/sdc/queue/nomerges

Increase requests for block queue

echo "1024" > /sys/block/sdc/queue/nr_requests

Example: Configure simple iSCSI Target with one portal and two LUNs

Assuming we have one iSCSI Target server with portal at, two LUNs (Malloc0 and Malloc), and accepting initiators on, like on diagram below:

Sample iSCSI configuration

Configure iSCSI Target

Start iscsi_tgt application:

$ ./app/iscsi_tgt/iscsi_tgt

Construct two 64MB Malloc block devices with 512B sector size "Malloc0" and "Malloc1":

$ python ./scripts/ construct_malloc_bdev -b Malloc0 64 512
$ python ./scripts/ construct_malloc_bdev -b Malloc1 64 512

Create new portal group with id 1, and address

$ python ./scripts/ add_portal_group 1

Create one initiator group with id 2 to accept any connection from

$ python ./scripts/ add_initiator_group 2 ANY

Finaly construct one target using previously created bdevs as LUN0 (Malloc0) and LUN1 (Malloc1) with a name "disk1" and alias "Data Disk1" using portal group 1 and initiator group 2.

$ python ./scripts/ construct_target_node disk1 "Data Disk1" "Malloc0:0 Malloc1:1" 1:2 64 -d

Configure initiator

Discover target

$ iscsiadm -m discovery -t sendtargets -p,1

Connect to the target

$ iscsiadm -m node --login

At this point the iSCSI target should show up as SCSI disks.

Check dmesg to see what they came up as. In this example it can look like below:

[630111.860078] scsi host68: iSCSI Initiator over TCP/IP
[630112.124743] scsi 68:0:0:0: Direct-Access INTEL Malloc disk 0001 PQ: 0 ANSI: 5
[630112.125445] sd 68:0:0:0: [sdd] 131072 512-byte logical blocks: (67.1 MB/64.0 MiB)
[630112.125468] sd 68:0:0:0: Attached scsi generic sg3 type 0
[630112.125926] sd 68:0:0:0: [sdd] Write Protect is off
[630112.125934] sd 68:0:0:0: [sdd] Mode Sense: 83 00 00 08
[630112.126049] sd 68:0:0:0: [sdd] Write cache: enabled, read cache: disabled, doesn't support DPO or FUA
[630112.126483] scsi 68:0:0:1: Direct-Access INTEL Malloc disk 0001 PQ: 0 ANSI: 5
[630112.127096] sd 68:0:0:1: Attached scsi generic sg4 type 0
[630112.127143] sd 68:0:0:1: [sde] 131072 512-byte logical blocks: (67.1 MB/64.0 MiB)
[630112.127566] sd 68:0:0:1: [sde] Write Protect is off
[630112.127573] sd 68:0:0:1: [sde] Mode Sense: 83 00 00 08
[630112.127728] sd 68:0:0:1: [sde] Write cache: enabled, read cache: disabled, doesn't support DPO or FUA
[630112.128246] sd 68:0:0:0: [sdd] Attached SCSI disk
[630112.129789] sd 68:0:0:1: [sde] Attached SCSI disk

You may also use simple bash command to find /dev/sdX nodes for each iSCSI LUN in all logged iSCSI sessions:

$ iscsiadm -m session -P 3 | grep "Attached scsi disk" | awk '{print $4}'

Vector Packet Processing

VPP (part of Fast Data - Input/Output project) is an extensible userspace framework providing networking functionality. It is build on idea of packet processing graph (see What is VPP?).

A detailed instructions for simplified steps 1-3 below, can be found on VPP Quick Start Guide.

SPDK supports VPP version 18.01.1.

1. Building VPP (optional)

Please skip this step if using already built packages.

Clone and checkout VPP

git clone && cd vpp
git checkout v18.01.1

Install VPP build dependencies

make install-dep

Build and create .rpm packages

make pkg-rpm

Alternatively, build and create .deb packages

make pkg-deb

Packages can be found in vpp/build-root/ directory.

For more in depth instructions please see Building section in VPP documentation

Please note: VPP 18.01.1 does not support OpenSSL 1.1. It is suggested to install a compatibility package for compilation time.

sudo dnf install -y --allowerasing compat-openssl10-devel

Then reinstall latest OpenSSL devel package:

sudo dnf install -y --allowerasing openssl-devel

2. Installing VPP

Packages can be installed from distribution repository or built in previous step. Minimal set of packages consists of vpp, vpp-lib and vpp-devel.

Note: Please remove or modify /etc/sysctl.d/80-vpp.conf file with appropriate values dependent on number of hugepages that will be used on system.

3. Running VPP

VPP takes over any network interfaces that were bound to userspace driver, for details please see DPDK guide on Binding and Unbinding Network Ports to/from the Kernel Modules.

VPP is installed as service and disabled by default. To start VPP with default config:

sudo systemctl start vpp

Alternatively, use vpp binary directly

sudo vpp unix {cli-listen /run/vpp/cli.sock}

A usefull tool is vppctl, that allows to control running VPP instance. Either by entering VPP configuration prompt

sudo vppctl

Or, by sending single command directly. For example to display interfaces within VPP:

sudo vppctl show interface

Example: Tap interfaces on single host

For functional test purpose a virtual tap interface can be created, so no additional network hardware is required. This will allow network communication between SPDK iSCSI target using VPP end of tap and kernel iSCSI initiator using the kernel part of tap. A single host is used in this scenario.

Create tap interface via VPP

vppctl tap connect tap0
vppctl set interface state tapcli-0 up
vppctl set interface ip address tapcli-0
vppctl show int addr

Assign address on kernel interface

sudo ip addr add dev tap0
sudo ip link set tap0 up

To verify connectivity


4. Building SPDK with VPP

Support for VPP can be built into SPDK by using configuration option.

configure --with-vpp

Alternatively, directory with built libraries can be pointed at and will be used for compilation instead of installed packages.

configure --with-vpp=/path/to/vpp/repo/build-root/vpp

5. Running SPDK with VPP

VPP application has to be started before SPDK iSCSI target, in order to enable usage of network interfaces. After SPDK iSCSI target initialization finishes, interfaces configured within VPP will be available to be configured as portal addresses. Please refer to Configuring iSCSI Target via RPC method.

iSCSI Hotplug

At the iSCSI level, we provide the following support for Hotplug:

  1. bdev/nvme: At the bdev/nvme level, we start one hotplug monitor which will call spdk_nvme_probe() periodically to get the hotplug events. We provide the private attach_cb and remove_cb for spdk_nvme_probe(). For the attach_cb, we will create the block device base on the NVMe device attached, and for the remove_cb, we will unregister the block device, which will also notify the upper level stack (for iSCSI target, the upper level stack is scsi/lun) to handle the hot-remove event.
  2. scsi/lun: When the LUN receive the hot-remove notification from block device layer, the LUN will be marked as removed, and all the IOs after this point will return with check condition status. Then the LUN starts one poller which will wait for all the commands which have already been submitted to block device to return back; after all the commands return back, the LUN will be deleted.

Known bugs and limitations

For write command, if you want to test hotplug with write command which will cause r2t, for example 1M size IO, it will crash the iscsi tgt. For read command, if you want to test hotplug with large read IO, for example 1M size IO, it will probably crash the iscsi tgt.

See also