Attach Linux, Windows and VMware hosts (multipath)

Part A checks the DS8000 side in the practice shell, and Part B builds a real two-path Linux setup in VirtualBox. A laptop has no Fibre Channel, so the VM uses two iSCSI paths (one per host-only NIC) to a LIO target on the same VM, and because the target reports vendor IBM and product 2107900, multipath treats the LUN like a DS8000 volume.

Practice shell note: Part A runs in our practice shell, which isn’t IBM software, so the output is real but a real DS CLI looks slightly different, and the SIM messages come from the shell. Part B runs on your own Ubuntu VM. We captured the multipath -t and package outputs on Ubuntu 24.04 ourselves, while VM outputs marked “something like this” are examples and yours will have different names, numbers and WWIDs.

1Part A: check the paths on the DS8000

Run this (practice shell, after setup-14.txt):

showhost erp_db1
lshostport -login

You should see this:

dscli> showhost erp_db1
Date/Time: 24 October 2026 14:09:40 UTC IBM DSCLI Version: 7.10.10.348 DS: IBM.2107-78SV001
Name          erp_db1
Type          Linux Server
State         Online
AddrMode      scsimap256
Volumes       8
Host ports    2
I/O ports     16 (all)
AddrDiscovery lunpolling
LBS           512
Cluster       erp_cluster
dscli> lshostport -login
Date/Time: 24 October 2026 14:09:40 UTC IBM DSCLI Version: 7.10.10.348 DS: IBM.2107-78SV001
WWNN             WWPN             Host    I/O ports ESSIOport
=============================================================
20000024FF2D0F86 21000024FF2D0F86 erp_db1 16 (all)  I0032
20000024FF2D0F87 21000024FF2D0F87 erp_db1 16 (all)  I0042
20000120FA13B914 10000090FA13B914 erp_db2 16 (all)  I0032
20000120FA13B915 10000090FA13B915 erp_db2 16 (all)  I0042

erp_db1 has two host ports, one logged in on I0032 (I/O enclosure 1B2) and one on I0042 (also 1B2 in this small lab box), whereas on a real build you’d want them in two different enclosures. With 8 volumes and 2 paths each, the host sees 16 SCSI devices before multipath gets involved.

✅ Checkpoint: Each ERP host has exactly two logged-in WWPNs.

2Part B: prepare saat-host1

In VirtualBox, give the VM two host-only adapters, say on networks 192.168.56.0/24 and 192.168.57.0/24, set the addresses 192.168.56.10 and 192.168.57.10 in netplan and then install the packages.

Run this (on saat-host1):

sudo apt update
sudo apt install -y targetcli-fb open-iscsi multipath-tools lsscsi
dpkg -l multipath-tools open-iscsi targetcli-fb | tail -3

You should see this (captured on Ubuntu 24.04):

$ dpkg -l multipath-tools open-iscsi targetcli-fb | tail -3
ii  multipath-tools 0.9.4-5ubuntu8.2  amd64        maintain multipath block device access
ii  open-iscsi      2.1.9-3ubuntu5.5  amd64        iSCSI initiator tools
ii  targetcli-fb    1:2.1.53-1ubuntu3 all          Command shell for managing the Linux LI

✅ Checkpoint: multipath-tools 0.9.4 and open-iscsi 2.1.9 are installed.

3Create a LUN that looks like a DS8000 volume

Create a 2 GiB file-backed LUN and set its SCSI vendor and product before you export it, because LIO won’t let you change them once the LUN is in use.

Run this:

sudo mkdir -p /var/lib/iscsi_disks
sudo targetcli /backstores/fileio create name=ds8k_lun0 file_or_dev=/var/lib/iscsi_disks/lun0.img size=2G
WWN=$(sudo find /sys/kernel/config/target/core -path '*ds8k_lun0/wwn' -type d)
echo IBM | sudo tee $WWN/vendor_id
echo 2107900 | sudo tee $WWN/product_id

You should see something like this (from a VM, not captured by us):

Created fileio ds8k_lun0 with size 2147483648
IBM
2107900

✅ Checkpoint: tee echoed IBM and 2107900 without “Invalid argument”.

4Publish it on two iSCSI portals

Run this:

T=iqn.2026-10.com.saatvedha:ds8k-lab
sudo targetcli /iscsi create $T
sudo targetcli /iscsi/$T/tpg1/luns create /backstores/fileio/ds8k_lun0
sudo targetcli /iscsi/$T/tpg1/portals delete 0.0.0.0 3260
sudo targetcli /iscsi/$T/tpg1/portals create 192.168.56.10 3260
sudo targetcli /iscsi/$T/tpg1/portals create 192.168.57.10 3260
INIT=$(sudo awk -F= '/^InitiatorName/{print $2}' /etc/iscsi/initiatorname.iscsi)
sudo targetcli /iscsi/$T/tpg1/acls create $INIT
sudo targetcli saveconfig

The ACL does the job of the DS8000 host mapping here, so only this initiator name can see the LUN.

✅ Checkpoint: sudo targetcli ls shows one LUN, two portals and one ACL.

5Log in over both paths

Start by shortening the iSCSI replacement timeout, so multipath can fail over in a few seconds instead of waiting the default 120 seconds, and then discover and log in to both portals.

Run this:

sudo sed -i 's/^node.session.timeo.replacement_timeout = 120/node.session.timeo.replacement_timeout = 15/' /etc/iscsi/iscsid.conf
sudo iscsiadm -m discovery -t sendtargets -p 192.168.56.10
sudo iscsiadm -m discovery -t sendtargets -p 192.168.57.10
sudo iscsiadm -m node --login
lsscsi | grep 2107900

You should see something like this (from a VM, not captured by us):

[2:0:0:0]    disk    IBM      2107900          4.0   /dev/sdb
[3:0:0:0]    disk    IBM      2107900          4.0   /dev/sdc

You get two SCSI disks for the same LUN, one per path, which is the same situation as a server zoned to two DS8000 ports.

✅ Checkpoint: Two IBM 2107900 disks appear.

6See the built-in DS8000 rule and turn multipath on

Run this:

multipath -t | grep -B2 -A4 2107900
sudo tee /etc/multipath.conf <<'EOF'
defaults {
    user_friendly_names yes
    find_multipaths     yes
}
EOF
sudo systemctl restart multipathd
sudo multipath -ll

You should see this for the first command (captured on Ubuntu 24.04):

$ multipath -t | grep -B2 -A4 2107900
	device {
		vendor "IBM"
		product "^2107900"
		path_grouping_policy "group_by_prio"
		failback "immediate"
		no_path_retry "queue"
	}

And something like this for multipath -ll (from a VM, not captured by us):

mpatha (36001405b3c3a1f2e4d6b8a9c0d1e2f30) dm-0 IBM,2107900
size=2.0G features='1 queue_if_no_path' hwhandler='1 alua' wp=rw
`-+- policy='service-time 0' prio=50 status=active
  |- 2:0:0:0 sdb 8:16 active ready running
  `- 3:0:0:0 sdc 8:32 active ready running

There’s one device, mpatha, with two healthy paths, and queue_if_no_path comes from the IBM rule (no_path_retry queue). Your WWID won’t match ours, since LIO uses its own 6001405 prefix while a real DS8000 LUN starts with 6005076.

✅ Checkpoint: multipath -ll lists one map with two “active ready running” paths.

7Fail one path while writing

Both portals are on the same VM, so pulling a virtual cable won’t break anything because the traffic never leaves the VM. Block one portal with a firewall rule instead, while a write runs in a second terminal.

Run this:

sudo mkfs.ext4 -q /dev/mapper/mpatha && sudo mount /dev/mapper/mpatha /mnt
sudo dd if=/dev/zero of=/mnt/test bs=1M count=1500 oflag=direct &
sudo iptables -I INPUT -d 192.168.57.10 -p tcp --dport 3260 -j DROP
sleep 30; sudo multipath -ll
sudo iptables -D INPUT -d 192.168.57.10 -p tcp --dport 3260 -j DROP

You should see something like this (from a VM, not captured by us):

mpatha (36001405b3c3a1f2e4d6b8a9c0d1e2f30) dm-0 IBM,2107900
size=2.0G features='1 queue_if_no_path' hwhandler='1 alua' wp=rw
`-+- policy='service-time 0' prio=50 status=active
  |- 2:0:0:0 sdb 8:16 active ready running
  `- 3:0:0:0 sdc 8:32 failed faulty running

dd keeps writing through sdb, and once you delete the rule, iscsid reconnects and the path is back to active within a minute.

✅ Checkpoint: One path failed, the write finished, and the path came back.