Updated: 28 September 2026 · Applies to: iscsi-initiator-utils 6.2.1 on AlmaLinux and Rocky Linux 9 and 10 and RHEL 9 and 10; open-iscsi on Ubuntu 24.04 and 26.04 and Debian 13
On a server with two or more network interfaces, iscsiadm -m discovery can hang and then fail with a timeout, although the storage answers on one of the interfaces. The two usual causes are that the kernel sends the request out of the wrong interface, and that reverse path filtering drops the replies when several interfaces are in the same subnet. The fix is to bind iSCSI to specific interfaces and, if needed, relax the filter.
1. Check the path to the storage
ip route get 10.10.10.50 ping -c 3 -I eth1 10.10.10.50
The first command shows which interface the kernel would use to reach the storage portal; the second tests one interface explicitly.
2. Bind iSCSI to the storage interfaces
Create one iSCSI interface record per network interface that should carry storage traffic:
iscsiadm -m iface -I iface-eth1 --op=new iscsiadm -m iface -I iface-eth1 --op=update -n iface.net_ifacename -v eth1 iscsiadm -m iface -I iface-eth2 --op=new iscsiadm -m iface -I iface-eth2 --op=update -n iface.net_ifacename -v eth2
Then discover and log in through those interfaces:
iscsiadm -m discovery -t sendtargets -p 10.10.10.50 -I iface-eth1 -I iface-eth2 iscsiadm -m node -l
With two bound interfaces you get two sessions per target, which multipath combines into one device (see how to scan for new LUNs with multipath).
3. Interfaces in the same subnet: reverse path filtering
If both storage interfaces have addresses in the same subnet, strict reverse path filtering (rp_filter = 1) can drop replies arriving on the "other" interface. Switch the storage interfaces to loose mode:
cat > /etc/sysctl.d/90-iscsi-rpfilter.conf <<'EOF' net.ipv4.conf.eth1.rp_filter = 2 net.ipv4.conf.eth2.rp_filter = 2 EOF sysctl --system
Separate subnets per storage path avoid this problem altogether and are the common design for iSCSI multipath.
4. Shorter timeouts for unreachable portals
If some listed portals are not reachable from this server, discovery waits for each. The login timeout can be lowered in /etc/iscsi/iscsid.conf with node.conn[0].timeo.login_timeout (default 15 seconds); restart iscsid afterwards.
Official documentation: Configuring an iSCSI initiator (Red Hat) and open-iscsi.
