Operating System - HP-UX
1838614 Members
3709 Online
110128 Solutions
New Discussion

Re: ServiceGuard and heartbeat subnet

 
Ermakov
New Member

ServiceGuard and heartbeat subnet

Hi
I have a trouble with my ServiceGuard 11.16 configuration (HP-UX) by ServiceGuard Manager(windows).
I prepare two nodes with shared storage and public lan interfaces in public subnet. I wrote hostnames and ip addr in etc/cmcluster/cmclnodelist file for initial configuration .... and i saw my two hosts in ServiceGuard Manager ready to be a cluster. But when I had configured my heartbeat LAN subnet (server-to server) i couldn't connect to hosts in ServiceGuard Manager and this error was displayed:
"No clusters were found or
the user does not have the
required permissions
Session disconnected"
I tried to write all hostnames and ip addr to etc/cmcluster/cmclnodelist but it didn't help me.
When i remove heartbeat subnet and restart one server it works well.
Will you help me, please.
6 REPLIES 6
RAC_1
Honored Contributor

Re: ServiceGuard and heartbeat subnet

/etc/cmcluster/cmclnodelist is not supported on this version of MCSG. You need to have those users in package configuration files itself.

Anil
There is no substitute to HARDWORK
D Block 2
Respected Contributor

Re: ServiceGuard and heartbeat subnet

can you do the linkloop command on hp-ux node1 to the node2's lan cards ?

see: lanscan, and linkloop

node1:
linkoop -i -v

and on node2, do the linkloop back to node1's linkaddr.
Golf is a Good Walk Spoiled, Mark Twain.
Ermakov
New Member

Re: ServiceGuard and heartbeat subnet

Ok. I have two interfaces on each server.
node1
lan0: 10.2.1.2 mask 255.255.254.0
lan1: 172.31.77.1 mask 255.255.255.0

node2
lan0: 10.2.1.3 mask 255.255.254.0
lan1: 172.31.77.2 mask 255.255.255.0

Default GW: 10.2.1.1

lan0 on each server plug into public network.
and node1\lan1 plug into node2\lan1 - its hearbeat subnet.
Linkloop shows that no connectivity from node1\lan0 to node2\lan1 and node1\lan1 to node2\lan0.
There is connectivity between public interfaces and private interfaces.
Route table is correct and ip (icmp) connectivity too.

Sincerely,
Georgy Ermakov.
Greg OBarr
Regular Advisor

Re: ServiceGuard and heartbeat subnet

You say your config is like this:

node1
lan0: 10.2.1.2 mask 255.255.254.0
lan1: 172.31.77.1 mask 255.255.255.0

node2
lan0: 10.2.1.3 mask 255.255.254.0
lan1: 172.31.77.2 mask 255.255.255.0

Default GW: 10.2.1.1

You don't need (and shouldn't have) a default gateway defined for 10.2.1.1. You need to be using a crossover cable if you're going directly out of lan0 on one box and into lan0 on the other box. If you don't have a crossover cable, you need to plug both into a little mini-hub, or configure a private vlan on your switch for the 10.2.1 subnet. That will work too. But since they're both on the same subnet, you don't need the default gateway defined for 10.2.1.1.

Here's how mine is setup:

node1
lan0: 10.3.3.1 mask 255.255.255.0 (h-beat)
lan2: 172.28.60.11 mask 255.255.255.0

node2
lan0: 10.3.3.2 mask 255.255.255.0 (h-beat)
lan2: 172.28.60.12 mask 255.255.255.0

and the default gateway is 172.28.60.1. I have a private vlan setup on the cisco switch for the 10.3.3 subnet that is only accessible by those two ports that I use for the heartbeat. (I'm using the switch instead of a crossover cable because the servers are located in separate remote areas of the facility for DR purposes). It works great.
Ermakov
New Member

Re: ServiceGuard and heartbeat subnet

I have solved a question: i used two dns servers in my resolv.conf.
xyz.ru

10.2.1.4
10.2.1.5

and i use hosts file too.
There were no problems in nslookup and name resolution on each box. i removed dns servers from my resolv.conf and started cluster successfuly. But a dont know why is it so. May be troubles with FQDN and ServiceGuard access policy.... i dont know.

Georgy
Stephen Doud
Honored Contributor

Re: ServiceGuard and heartbeat subnet

cmclnodelist IS supported with A.11.16 before building the first cluster binary file (see http://docs.hp.com/en/B3936-90079/ch05s01.html - Preparing your system)
"With A.11.16, if a node is not configured in a cluster, Serviceguard will check for an /etc/cmcluster/cmclnodelist file. "

Due to new security enhancements to Serviceguard, /etc/nsswitch.conf should be configured to look at /etc/hosts before looking at DNS server.

# grep hosts /etc/nsswitch.conf
hosts: files [NOTFOUND=continue] dns

The way in which /etc/hosts should be configured is documented in the same section of the document I linked above, in the section titled:
"Name Resolution Following Primary LAN Failure or Loss of DNS"