2 minute delay when booting over iSCSI

Issues related to configuring your network
Post Reply
jem
Posts: 1
Joined: 2020/11/04 14:39:22

2 minute delay when booting over iSCSI

Post by jem » 2020/11/04 14:46:18

Hi all.

I've recently built three diskless CentOS 8 hosts that boot using iSCSI. In addition to their onboard network interfaces, they are each equipped with a dual-port Intel I350-T2 NIC that has had the PXE firmware replaced with iSCSI Remote Boot firmware. The interfaces have the following purposes:

Code: Select all

Interface	        VLAN	                                IP subnet
eno1 (onboard)	        VLAN 128 (general connectivity)	        10.0.128.x/24
enp1s0f0 (I350 port 1)	VLAN 192 (dedicated iSCSI traffic)	10.0.192.x/24
enp1s0f1 (I350 port 2)	VLAN 193 (dedicated iSCSI traffic)	10.0.193.x/24
The iSCSI storage server presents each target on both VLAN 192 and 193 and the iSCSI NICs are configured to connect to it at boot. When I run the CentOS installer, it automatically detects the target LUNs as multipath devices and I can perform a successful installation to them. So far so good.

However, when I boot up the hosts, they all encounter a two minute delay where nothing appears to be happening. The last thing on the console is:

Code: Select all

dracut-initqueue[x]: RTNETLINK answers: File exists
scsi 6:0:0:0: alua: rtpg retry
scsi 6:0:0:1: alua: rtpg retry
At that point, if I run a packet capture on those VLANs, I can see the host is making ARP 'who-has' requests for its VLAN 128 gateway (10.0.128.1) every second for a minute, but it's sending those requests on its VLAN 192 interface, where it never gets an answer. It then does the same thing on its VLAN 193 interface for another minute. After that, the boot continues normally.

Code: Select all

ARP, Request who-has 10.0.128.1 (ff:ff:ff:ff:ff:ff) tell 10.0.192.53, length 46
(repeated 60 times)

ARP, Request who-has 10.0.128.1 (ff:ff:ff:ff:ff:ff) tell 10.0.193.53, length 46
(repeated 60 times)
All interfaces are getting their IP addresses by DHCP, but no gateway is included in the VLAN 192 and 193 DHCP responses, as those are isolated subnets that don't need to route traffic anywhere else.

Does anyone have any suggestions for what might be going on here? Why is it ARPing for an IP address from interfaces that are in a different IP subnet?

Thanks.

Post Reply

Return to “CentOS 8 - Networking Support”