Unsolved
This post is more than 5 years old
2 Posts
0
12535
June 3rd, 2014 13:00
Md3200i ports refusing connections
Hi,
After some months running smoothly, an MD3200i storage array (brand new, less than 6 months) has started to show an strange behaviour. Despite iscsiadm showing all the 8 ports in discovery mode (multipathing with 2 nic cards), fails to login in 2 specific portals. Always the same two, no matter which host you try to login from. It was working before. Reviewing iscsiadm logs, we saw:
Jun 3 17:13:24 akira iscsid: connect to 10.200.10.7:3260 failed (Connection refused)
Jun 3 17:13:24 akira iscsid: connect to 10.200.10.2:3260 failed (Connection refused)
Checking 3260 port in the storage array shows it closed, while the rest of the ports show it open. Is this normal? I've been making searches in Google and found no traces of something like this.
Can someone help?
Thanks,
Rubén.
Dev Mgr
4 Operator
•
9.3K Posts
0
June 3rd, 2014 15:00
Are you using 4 subnets like you are supposed to?
Can you ping the 2 ports on the SAN?
Ruben Cardenal
2 Posts
0
June 3rd, 2014 16:00
Hi,
I'm aware of the recommended settings and no, I'm using a single /24 (10.200.10.0) on a dedicated vlan, we can't afford 4 switches and/or (in case we use 4 vlans) 4 dedicated nics per host because it would make cabling crazy there, so 2 ports from each server to the iscsi vlan (they have 4, we need the rest for other stuff). We'll expand that topology in the future.I have this setup in other storages of this kind and works well. Not RFCish? Maybe, but works. Being polite here :)
Anyway using a different addressing scheme doesn't change the fact that those two ports have the iscsi protocolt port closed, when it should be opened as it was before, and like the rest of the ports are. Yes, they're pingable.
On an example server, having eth2 (10.200.10.12) and eth3 (10.200.10.13) and the iscsi ports 10.200.10.1-8, nc shows:
# for f in 12 13; do for g in $(seq 1 8); do nc -z -s 10.200.10.$f 10.200.10.$g 3260 ; echo " 10.200.10.$f -> 10.200.10.$g = $?"; done; done
10.200.10.12 -> 10.200.10.1 = 0
10.200.10.12 -> 10.200.10.2 = 1
10.200.10.12 -> 10.200.10.3 = 0
10.200.10.12 -> 10.200.10.4 = 0
10.200.10.12 -> 10.200.10.5 = 0
10.200.10.12 -> 10.200.10.6 = 0
10.200.10.12 -> 10.200.10.7 = 1
10.200.10.12 -> 10.200.10.8 = 0
10.200.10.13 -> 10.200.10.1 = 0
10.200.10.13 -> 10.200.10.2 = 1
10.200.10.13 -> 10.200.10.3 = 0
10.200.10.13 -> 10.200.10.4 = 0
10.200.10.13 -> 10.200.10.5 = 0
10.200.10.13 -> 10.200.10.6 = 0
10.200.10.13 -> 10.200.10.7 = 1
10.200.10.13 -> 10.200.10.8 = 0
The same repeats all over the rest of the servers attached to the storage (4 at this time): ports 011 and 112 refuse connections. The MD Storage Manager shows no errors. There aren't any erros in the nics and in the switch ports. It's just that... port closed in those cases.
Rubén.