Error: Detected a partition of IP subnet

[Omar Alvi_1]

Hi,

I'm configuring a simple 2 node cluster. I have 2 network cards on each system, with each set of interfaces configured on a different subnet.

Subnet A: 10.10.30.0
Subnet Mask: 255.255.255.0

Subnet B: 100.100.0.0
Subnet Mask: 255.255.0.0

The second subnet is isolated, as in two network cables connected to an isolated hub.

All the linkloop outputs between each of the interfaces is fine ... even checked using cmscancl.

But still when I run cmcheckconf I get the following errors

Error: Detected a partition of IP subnet 100.100.0.0.
Partition 1
hp11 lan1
Partition 2
sux4 lan1
Failed to evaluate network
cmcheckconf : Unable to reconcile configuration file cmclconfig.ascii
with discovered configuration information.

How can this error be resolved, and what does it mean in the first place.

Would appreciate any assistance.

ServiceGuard Version is 11.16 on HP-UX 11i V1

Thanks and Regards,

-Alvi

[melvyn burnard]

essentially, it is telling you that there is something that causes SG to believe this network is not connected.
I would run a cmquerycl again, writing it to a test file, and compare the network information that gets put in the asciifile.
Also, check that something is not stopping tcp communication going through the network.
You could post the result of ifconfig lan1 from each node here, also lanscan, and maybe the node info that is in the ascii file

[Mel Burslan]

did you actually do a linkloop from hp11 lan1 to sux4 lan1 mac address and the reverse direction ? Sometimes when people say they did linkloop tests, they test the local interface.

All in all this what you have said points to the fact that the two nodes can not communicate overl lan1 due to a some cause. Is it a plain hub connecting them or a switch. If it is a switch, are the two ports these lan1s are connected on the same vlan ? Or are you just using a crossover cable, in which case is it wired correctly ?

[Omar Alvi_1]

Thanks for the replies,

I was able to resolve the intiial problem; it turned out that the lans were totally isolated.

With the route add host command and making use of hp-ux's inherent routing capabilities I got the lans to communicate.

Just one issue. What is the point eventually

A- That each lan1 , aln2 and lan3 can communicate with the other respectively

or

B- Total interconnectivity where each lan can linkloop to each and every other lan whether it be lan0, 1 or 2

If its B, its a tough task as private networks and vlans are security requirements.

Any assistance is appreciated.

Thanks and Regards,

-Alvi

[melvyn burnard · Accepted solution]

you do not HAVE to hav elan1, lan2 and lan3 all communication to each other, but iof you wish to use a lan as a standby for a configured lan in Serviceguard, they MUST be able to communicate with each other out on the physical network, otherwise how can the standby then act as th eworking lan when the primary fails.

[Omar Alvi_1]

I have the following scenario

Systam A and Systm B each having three lans on differents subnets and vlans.

I have a downtime soon and I was preparing the prerequisites for installation.

Following is the cmscancl linkloop output from one system to the other

###### Checking REMOTE network connections (systemB to systemA) ######

------ lan0 on node systemB to lan0 on node systemA ------
PPA 0 link test to 0x000F202B4550 -- OK

------ lan0 on node systemB to lan1 on node systemA ------
PPA 0 link test to 0x00306EF56593 (NO CONNECTION)

------ lan0 on node systemB to lan2 on node systemA ------
PPA 0 link test to 0x00306EF565A3 (NO CONNECTION)

------ lan1 on node systemB to lan0 on node systemA ------
PPA 1 link test to 0x000F202B4550 (NO CONNECTION)

------ lan1 on node systemB to lan1 on node systemA ------
PPA 1 link test to 0x00306EF56593 -- OK

------ lan1 on node systemB to lan2 on node systemA ------
PPA 1 link test to 0x00306EF565A3 (NO CONNECTION)

------ lan2 on node systemB to lan0 on node systemA ------
PPA 2 link test to 0x000F202B4550 (NO CONNECTION)

------ lan2 on node systemB to lan1 on node systemA ------
PPA 2 link test to 0x00306EF56593 (NO CONNECTION)

------ lan2 on node systemB to lan2 on node systemA ------
PPA 2 link test to 0x00306EF565A3 â OK


From my point of view this is acceptable as I have no standby lans, all these are active lans

As for the linkloop within the system none of the lans can communicate with each other as they're on different vlans

------ lan0 to lan1 ------
PPA 0 link test to 0x00306EF56593 (NO CONNECTION)

------ lan0 to lan2 ------
PPA 0 link test to 0x00306EF565A3 (NO CONNECTION)

------ lan1 to lan0 ------
PPA 1 link test to 0x000F202B4550 (NO CONNECTION)

------ lan1 to lan2 ------
PPA 1 link test to 0x00306EF565A3 (NO CONNECTION)

------ lan2 to lan0 ------
PPA 2 link test to 0x000F202B4550 (NO CONNECTION)

------ lan2 to lan1 ------
PPA 2 link test to 0x00306EF56593 (NO CONNECTION)


Isn't this acceptable from MCSG point of view.

If not then what alternatives do I have? Can't reconfigure the vlans.

Thanks and Regards,

-Alvi

[melvyn burnard]

Yes, if you only wish to have active lans, then this is acceptable. Just make sure all three lans are HEARTBEAT_IP in the cluster ascii file.

To check if all is ok, use the cmquerycl comman and check the ascii file against what you expect