Well, the hosts-file is not maintained (since yeah, they have naming services running^^). IP Config in terms of name and domain is set up likewise on all servers, so hostname XYZ, domain de.xyz.abc.com, primary DNS suffix identical, primary and secondary DNS servers identical, no WINS. They have a stringent deployment procedure for their physical servers that doesn't deviate from the defined standard on individual hosts. So there's no differences among hosts which could render this behaviour.
Are you going for something similar as with Unisphere Host Agent and the agentid.txt file to make sure a server registers properly in a VNX?
exactly, as if it's "binding" to the wrong NIC or something ? There should be a place in TCP configuration where you can look and binding, can't remember from the top of my head. Interesting to check if that differs between the hosts.
Yeah, they have. They're actually HP C7000 Blades (Gen8) with 2 dual port Emulex CNAs (but HP OEM stuff) that carry both Ethernet- and FCoE traffic. Each CNA is personalized as FCoE on one port and as NIC on the other. NICs do form a team, FCoE paths are Powerpath-protected.
Each server is equipped with a bunch of IP addresses, but Powerpath Viewer discovery is pointing at the DNS name of course: i have multiple .csv files in place for different datacenters to distinguish among groupwise in PP viewer, just containing the server names, not FQDN. I have tried to go for FQDN during discovery, but no change: PP viewer says "connected", but states either the NETBIOS name or the IP.
Are you going for something similar as with Unisphere Host Agent and the agentid.txt file to make sure a server registers properly in a VNX?
Might be worthwile to check, I agree. But I doubt that it's differently done on the nodes that are properly displayed vs. the screwed ones. I'll check next week when i'm back onsite. It's a bit complicated since the binding stuff is in this case seems to be covered by HP means (NCU) that also do the NIC teaming and so on. will get back to you once I figured it out.
Wolfi72
12 Posts
0
August 8th, 2014 04:00
Well, the hosts-file is not maintained (since yeah, they have naming services running^^). IP Config in terms of name and domain is set up likewise on all servers, so hostname XYZ, domain de.xyz.abc.com, primary DNS suffix identical, primary and secondary DNS servers identical, no WINS. They have a stringent deployment procedure for their physical servers that doesn't deviate from the defined standard on individual hosts. So there's no differences among hosts which could render this behaviour.
dynamox
9 Legend
•
20.4K Posts
0
August 8th, 2014 04:00
do you know if those servers that come in incorrectly, do they have multiple NICs ?
dynamox
9 Legend
•
20.4K Posts
0
August 8th, 2014 04:00
what about the hosts themselves, how is the host name configured there ? For example on Linux i would look in /etc/hosts, /etc/sysconfig/network
dynamox
9 Legend
•
20.4K Posts
0
August 8th, 2014 05:00
exactly, as if it's "binding" to the wrong NIC or something ? There should be a place in TCP configuration where you can look and binding, can't remember from the top of my head. Interesting to check if that differs between the hosts.
Wolfi72
12 Posts
0
August 8th, 2014 05:00
Yeah, they have. They're actually HP C7000 Blades (Gen8) with 2 dual port Emulex CNAs (but HP OEM stuff) that carry both Ethernet- and FCoE traffic. Each CNA is personalized as FCoE on one port and as NIC on the other. NICs do form a team, FCoE paths are Powerpath-protected.
Each server is equipped with a bunch of IP addresses, but Powerpath Viewer discovery is pointing at the DNS name of course: i have multiple .csv files in place for different datacenters to distinguish among groupwise in PP viewer, just containing the server names, not FQDN. I have tried to go for FQDN during discovery, but no change: PP viewer says "connected", but states either the NETBIOS name or the IP.
Are you going for something similar as with Unisphere Host Agent and the agentid.txt file to make sure a server registers properly in a VNX?
Wolfi72
12 Posts
0
August 8th, 2014 06:00
Might be worthwile to check, I agree. But I doubt that it's differently done on the nodes that are properly displayed vs. the screwed ones. I'll check next week when i'm back onsite. It's a bit complicated since the binding stuff is in this case seems to be covered by HP means (NCU) that also do the NIC teaming and so on. will get back to you once I figured it out.