HACMP/ES 5.1 + IPAT with IP alias problems
From: Goran Koruga (gorank_at_hermes.si)
Date: 01/23/04
- Next message: Dale Talcott: "Re: 64-bit kernel on AIX5.1 POWER3/"
- Previous message: steven: "aix license"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: 23 Jan 2004 15:43:16 +0100
Hi all.
I was hoping to find someone that could perhaps shed some light on the
issues I am having when trying to configure HACMP/ES 5.1 on a two node
cluster, with a very simple cluster configuration. I only run into this
when trying IPAT with IP aliases, I didn't have any issues when I used
IPAT with IP replacemenet.
I will try to describe my configuration in more details, hopefully
providing all the required info.
I have a two node cluster, cluster nodes are kolhoz and sovhoz. Each of
these has two network interfaces, and there's also a serial link between
the two. I have tried using various different network configurations,
but "HACMP Verification and Synchronization" SMIT screen always results
an error, no matter what I do. Here's the error message :
---------------------------------------------------------------
Verifying Cluster Topology...
ERROR: Service adapter sovhoz-svc on network ito_ether is on the same\n
subnet as at least one of the boot adapters on this network.\nService
labels must be on a different subnet when the network is configured\n to
use IP aliasing for IP address takeover.\nERROR: Service adapter
kolhoz-svc on network ito_ether is on the same\n subnet as at least one
of the boot adapters on this network.\nService labels must be on a
different subnet when the network is configured\n to use IP aliasing for
IP address takeover.\nERROR: Service adapter sovhoz-svc on network
ito_ether is on the same\n subnet as at least one of the boot adapters
on this network.\nService labels must be on a different subnet when the
network is configured\n to use IP aliasing for IP address
takeover.\nERROR: Service adapter kolhoz-svc on network ito_ether is on
the same\n subnet as at least one of the boot adapters on this
network.\nService labels must be on a different subnet when the network
is configured\n to use IP aliasing for IP address takeover.\n
Verifying Cluster Resources...
---------------------------------------------------------------
Here's how the node's networking is configured (it's very similar on the
other node, IP-s are just decreased by 1) :
---------------------------------------------------------------
# ifconfig -a
en0: flags=4e080863<UP,BROADCAST,NOTRAILERS,RUNNING,SIMPLEX,MULTICAST,GROUPRT,64BIT,PSEG>
inet 10.17.8.252 netmask 0xffff0000 broadcast 10.17.255.255
en1: flags=4e080863<UP,BROADCAST,NOTRAILERS,RUNNING,SIMPLEX,MULTICAST,GROUPRT,64BIT,PSEG>
inet 10.18.1.113 netmask 0xffff0000 broadcast 10.17.255.255
lo0: flags=e08084b<UP,BROADCAST,LOOPBACK,RUNNING,SIMPLEX,MULTICAST,GROUPRT,64BIT>
inet 127.0.0.1 netmask 0xff000000 broadcast 127.255.255.255
inet6 ::1/0
# netstat -nr
Routing tables
Destination Gateway Flags Refs Use If PMTU Exp Groups
Route tree for Protocol Family 2 (Internet):
default 10.17.250.250 UGc 0 0 en0 - -
10.17/16 10.17.8.252 U 3 2563 en0 - -
10.17.8.252 127.0.0.1 UGHS 4 1409 lo0 - -
10.18/16 10.18.1.11 U 0 174 en1 - -
10.18.1.11 127.0.0.1 UGHS 0 0 lo0 - -
127/8 127.0.0.1 U 4 193 lo0 - -
Route tree for Protocol Family 24 (Internet v6):
::1 ::1 UH 0 0 lo0 16896 -
---------------------------------------------------------------
To clarify :
10.17.8.252 sovhoz # boot
10.17.6.116 sovhoz-svc
10.18.1.15 sovhoz-svc18
10.18.1.11 sovhoz-stdby
Here's how my resource group is configured (the other one is pretty much
the same, but with reversed node order and different service address - a
normal mutual-takeover scenario) :
---------------------------------------------------------------
# /usr/es/sbin/cluster/utilities/clshowres -n'sovhoz'
Resource Group Name sovhoz_res_group
Node Relationship rotating
Site Relationship ignore
Participating Node Name(s) sovhoz kolhoz
Node Priority
Service IP Label sovhoz-svc
... (default settings come here)
---------------------------------------------------------------
Documentation lists some requirements related to IPAT with IP aliases,
depending on resource group configuration. However, it also says :
---------------------------------------------------------------
* Subnet requirements:
a) Multiple non-service addresses configured on a node should be
defined on different subnets.
b) Multiple service IP labels on the same node are not required to be
on different subnets.
c) If you are configuring a cascading resource group, the service
address must be on a different subnet from ALL non-service
addresses defined for that network on the cluster node. If you are
configuring a rotating resource group, the service IP label can be
on the same subnet as one of the non-service IP labels for that
network. This requirement enables HACMP to comply with the IP route
striping functionality of AIX 5L 5.1, which allows multiple routes
to the same subnet.
d) The netmask for all IP labels in an HACMP network must be the same.
---------------------------------------------------------------
I think I comply with all of the above. All my non-service addresses
are on different subnets (10.17 and 10.18), I use a rotating resource
group so the subnet for service IP address should not matter, and
finally, the netmask is the same for all IP labels. I tried using two
different service IP-s (kolhoz-svc and kolhoz-svc18), but this doesn't
help.
Can someone please point out if there are any mistakes in my
configuration ? Because to me it seems everything is configured as it
should be, and either documentation is wrong, or verification is not
taking the fact that my resource group is "rotating" so service vs.
non-service subnet should not matter.
Regards,
Goran
- Next message: Dale Talcott: "Re: 64-bit kernel on AIX5.1 POWER3/"
- Previous message: steven: "aix license"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
