1836451 Members
2389 Online
110100 Solutions
New Discussion

a question about MC/SG

 
SOLVED
Go to solution
??_29
Advisor

a question about MC/SG

Entironment:A rp8400 with two parition is configured MC/SG in a cluster.
I ping the heardbeat ip address and don't receive the echo package.But I press the command "cmruncl -v" ,two node both run.I think that they should not start because there is a communication failure of heartbeat between the nodes.Why they can run normally?
Thanks
Weit
12 REPLIES 12
Balaji N
Honored Contributor

Re: a question about MC/SG

hi ??,
i guess there is an alternate heart beat configured. it is using that. check for any errors / logs in the syslog.

-balaji
Its Always Important To Know, What People Think Of You. Then, Of Course, You Surprise Them By Giving More.
Armin Feller
Honored Contributor

Re: a question about MC/SG

How to you ping the lan interface? The heardbeat packages are send on layer 2 and you have to ping them with "linkloop" command. To check your MC/SG environment please use the "cmscancl" command, that creates a output with a lot of informations, also about your network.

# cmscancl >/tmp/cmscancl.out
# more /tmp/cmscancl.out

Regards ...
Armin
??_29
Advisor

Re: a question about MC/SG

hi
How can I find the information about alternate
heartbeat?Where is the information configured?
Thanks for your helping.
Armin Feller
Honored Contributor

Re: a question about MC/SG

Please have a look in your cluster ascii file.

# more /etc/cmcluster/cmclconf.ascii

Regards ...
Armin
Balaji N
Honored Contributor

Re: a question about MC/SG

check your ascii configuration file in /etc/cmcluster directory
-balaji
Its Always Important To Know, What People Think Of You. Then, Of Course, You Surprise Them By Giving More.
??_29
Advisor

Re: a question about MC/SG

Hi
Look Armin,this is the detail of the cluster.ascii file.

CLUSTER_NAME cluster1
FIRST_CLUSTER_LOCK_VG /dev/vglock
NODE_NAME ssokeysa
NETWORK_INTERFACE lan0
HEARTBEAT_IP 15.79.48.187
NETWORK_INTERFACE lan2
NETWORK_INTERFACE lan1
HEARTBEAT_IP 192.1.1.1
FIRST_CLUSTER_LOCK_PV /dev/dsk/c6t0d0

NODE_NAME ssokeysb
NETWORK_INTERFACE lan0
HEARTBEAT_IP 15.79.48.188
NETWORK_INTERFACE lan2
NETWORK_INTERFACE lan1
HEARTBEAT_IP 192.1.1.2
FIRST_CLUSTER_LOCK_PV /dev/dsk/c11t0d0

HEARTBEAT_INTERVAL 1000000
NODE_TIMEOUT 5000000
AUTO_START_TIMEOUT 600000000
NETWORK_POLLING_INTERVAL 2000000
MAX_CONFIGURED_PACKAGES 2

VOLUME_GROUP /dev/vglock
VOLUME_GROUP /dev/vgdata

The IP 192.1.1.1 to IP 192.1.1.2 don't communicate each other.
Why the cluster can run OK?


Armin Feller
Honored Contributor

Re: a question about MC/SG

Please run the "cmclscan" command like I have written. In the output you find all about the network connection in all directions, ...

Please paste the network part in the formen and I will try to explain.
??_29
Advisor

Re: a question about MC/SG

This is the scancl.doc.Thanks Armin.
BFA6
Respected Contributor

Re: a question about MC/SG

Hi,

Looking at your cluster.ascii file, you have the heartbeat going out on bothy lan cards, so even if the dedicated heartbeat lan does not work, a heartbeat is still getting through on the data lan.

Regards,

Hilary
Armin Feller
Honored Contributor

Re: a question about MC/SG

Hi,

in the following you can see that the lan connection between your nodes works fine on layer 2 (DLPI):

###### Checking REMOTE network connections (ssokeysb to ssokeysa) ######

------ lan0 on node ssokeysb to lan0 on node ssokeysa ------
PPA 0 link test to 0x00306E0AC052 -- OK

------ lan0 on node ssokeysb to lan1 on node ssokeysa ------
PPA 0 link test to 0x00306E21B56F (NO CONNECTION)

------ lan0 on node ssokeysb to lan2 on node ssokeysa ------
PPA 0 link test to 0x00306E21D516 -- OK

------ lan1 on node ssokeysb to lan0 on node ssokeysa ------
PPA 1 link test to 0x00306E0AC052 (NO CONNECTION)

------ lan1 on node ssokeysb to lan1 on node ssokeysa ------
PPA 1 link test to 0x00306E21B56F (NO CONNECTION)

------ lan1 on node ssokeysb to lan2 on node ssokeysa ------
PPA 1 link test to 0x00306E21D516 (NO CONNECTION)

------ lan2 on node ssokeysb to lan0 on node ssokeysa ------
PPA 2 link test to 0x00306E0AC052 -- OK

------ lan2 on node ssokeysb to lan1 on node ssokeysa ------
PPA 2 link test to 0x00306E21B56F (NO CONNECTION)

------ lan2 on node ssokeysb to lan2 on node ssokeysa ------
PPA 2 link test to 0x00306E21D516 -- OK


So if you are not able to ping (layer 3) between the nodes, please check the config of your switch/router/bridge between the nodes, if layer 3 (IP) is open.

In my mind and in view of MC/SG all works fine. MC/SG only check the lan on layer 2 and that seam to be ok.

Regards ...
Armin
??_29
Advisor

Re: a question about MC/SG

Can I explain that I have the heartbeat going out the both lan cards(lan0,lan1)?Lan 1 is dedicated heartbeat lan.If the lan1 is down,the heartbeat message can pass through the lan0.So the cluster can also run normally.
You can see the ssokeysa's lan1 to ssokeysb don't connect.
Thanks all of your help.I love this forum!:)
Armin Feller
Honored Contributor
Solution

Re: a question about MC/SG

There are 2 heardbeat connections between the nodes:

NODE_NAME ssokeysa
NETWORK_INTERFACE lan0 HEARTBEAT_IP 15.79.48.187
NETWORK_INTERFACE lan1 HEARTBEAT_IP 192.1.1.1

NODE_NAME ssokeysb
NETWORK_INTERFACE lan0 HEARTBEAT_IP 15.79.48.188
NETWORK_INTERFACE lan1 HEARTBEAT_IP 192.1.1.2

If on one node lan0 fails the heartbeat only go by lan1, for MC/SG no problem. DLPI pings can only work in the same physical network, so only lan0 to lan0 and lan1 to lan1.

------ lan0 on node ssokeysb to lan0 on node ssokeysa ------
PPA 0 link test to 0x00306E0AC052 -- OK

lan0 seams to be fine!!!

------ lan1 on node ssokeysb to lan1 on node ssokeysa ------
PPA 1 link test to 0x00306E21B56F (NO CONNECTION)

lan1 seams to be broken connection.

But like I'd said,if one of both heartbeat connections is working fine, MC/SG will start and protect your datas by corruption.

Regards ...
Armin