LAN card failure (SG A.11.19 environment)

[meekrob]

Good day All,

we have a SG A.11.19 environment of 2 rx7640 (HP-UX 11iv3) integrity nodes. One node appears to have one of its LAN card failed (NIC P/N: AD337A). The only card that we have available is a combo SCSI/2-port with P/N: AB290A

The question is can we replace the AD337A one (dual port) with the AB290A? and in this case do we have to modify anything in the cluster configuration files and if it is the case, which files do we need to modify?

Thanks in advance

[meekrob]

Hi Team, any help please is much appreciated.

I think that to reconfigure the NICs and ip addresses within the cluster, we 1st need to stop the cluster and then edit the cluster ascii file and once edited with the new NICs and IP addresses, run the cmapplyconf command.

Is it right? or do i still have to edit other config files?

Thanks in advance

[Matti_Kurkela]

You're right, re-applying the cluster configuration (after any necessary edits) should be the correct solution.

If the package configuration files contain any references to the particular NIC you might need to edit & re-apply them too, but I think the package configuration usually doesn't have any NIC names or hardware path references in them (because Serviceguard accepts that a particular subnet might be on lanX on one node and on lanY on another).

Essentially, if you have some custom scripting in the package configuration that depends on NIC names/hardware paths/MAC addresses, then you'll need to review it and possibly adjust it to match the new hardware configuration. But if you don't have anything like that, then you should have no need to re-apply the package configuration(s), only the cluster configuration.

[meekrob]

Many Thanks Matti for your precious help.

Just have one concern please:

consider that we have the below LAN config:

root@srvdb1:/> ioscan -fnC lan
Class     I  H/W Path        Driver S/W State   H/W Type     Description
=========================================================================
lan       0  0/0/1/1/0/6/0   iether   CLAIMED     INTERFACE    HP AB290-60001 PCI/PCI-X 1000Base-T 2-port U320 SCSI/2-port 1000B-T Combo Adapter
lan       1  0/0/1/1/0/6/1   iether   CLAIMED     INTERFACE    HP AB290-60001 PCI/PCI-X 1000Base-T 2-port U320 SCSI/2-port 1000B-T Combo Adapter
lan       2  1/0/1/1/0/6/0   iether   CLAIMED     INTERFACE    HP AB290-60001 PCI/PCI-X 1000Base-T 2-port U320 SCSI/2-port 1000B-T Combo Adapter
lan       3  1/0/1/1/0/6/1   iether   CLAIMED     INTERFACE    HP AB290-60001 PCI/PCI-X 1000Base-T 2-port U320 SCSI/2-port 1000B-T Combo Adapter
lan       4  1/0/12/0/0/0/0  iether   CLAIMED     INTERFACE    HP AD337-60001 PCIe 1000Base-T Dual-port Adapter
lan       5  1/0/12/0/0/0/1  iether   CLAIMED     INTERFACE    HP AD337-60001 PCIe 1000Base-T Dual-port Adapter

and in the cluster config we need to swap lans 4 & 5 ports that belong to a dual port NIC to another NIC having Lans 2 & 3 so in this case we just edit the cluster config ascii file and cmapplyconf? do we still need to edit other files as for example /etc/hosts  or /etc/rc.config.d/netconf and other files ...

Note that Lans 2 and 3 will have the same ip addresses that belonged to lans 4 & 5.

Thanks in advance

[Matti_Kurkela]

/etc/hosts only deals with hostnames associated with IP addresses. If you aren't adding any IP addresses you weren't using before, you should not need to touch that file.

But yes, /etc/rc.config.d/netconf will definitely need some changes too.

The sequence of operations should be like this:

• use cmhaltnode to halt all cluster operations on the node that will have its configuration modified
• if you plan to reboot this node (recommended to make sure all your changes are valid and the system is bootable with the new settings), edit /etc/rc.config.d/cmcluster to temporarily disable Serviceguard start-up on boot.
• change the local network configuration (/etc/rc.config.d/netconf, physical cabling changes if necessary)
• (remember: don't leave backup copies of any configuration files in /etc/rc.config.d directory, since all the files in this directory will be read by the system. The start-up scripts will automatically avoid a few common backup file suffixes, but it'll be safer to not rely on that.)
• at this point, it is recommended to reboot the node to verify that the system reboots with the new network configuration, but it isn't strictly necessary (alternative: make the changes using commands like ifconfig and route, and very very carefully check your /etc/rc.config.d/netconf changes for errors)
• get a copy of the current cluster configuration with cmgetconf (please don't use any old version of the cluster ASCII file you find laying around, unless you are sure it's up to date)
• modify the cluster configuration ASCII file and update it with cmapplyconf
• use cmrunnode to restart the cluster operations on this node
• check the syslog for any errors
• if everything is OK and you disabled Serviceguard start-up on boot, re-enable it again.

[meekrob]

Many Thanks Matti for your help

Crystal clear and convincing as usual ;)

Meanwhile, i will keep this thread open the time to schedule a maintenance window

Regards

[meekrob]

Good day Matti and all,

we ve scheduled a maintenance window end of the current week so just a quick review on the steps to proceed wiith as this environment is critical to our business.

Note that we need to perform same "LAN swapping" on both nodes of the cluster by switching LANs 4 and 5 to 2 and 3 respectively. LAN4 is supposed to be the standby for the primary and LAN5 is supposed to be the standby for the heartbeat. So as those steps need to be performed the same on both nodes => in this case we can halt the cluster and proceed as follows:

1) cd /etc/cmcluster

2) cmgetconf -v -c  <cluster_name>  temp.ascii

3) halt the cluster:  cmhaltcl -v

4) edit the temp.ascii file using vi editor

5) supposing that we do not need to edit /etc/rc.config.d/netconf  to reflect the new cabling changes as we are just disconnecting LAN4 and 5 cables and connecting them to LAN 2 and 3 and those LANs as they are used as Standby => no ip addresses assigned to these LANs.

6) cmcheckconf -v -C temp.ascii

7) cmapplyconf -v -C temp.ascii

8) start back the cluster with the new config using cmruncl -v

Kindly confirm please if it seems OK

Thanks in advance