Seiral Line as heartbeat in MCSG

[Christopher B. Santos]

Can anyone help me find docs on how to configure serial line of rp5470 server on 2 node MCSG? I can't find anything about configuring serial line as heartbeat.

[Isralyn Manalac_1]

Hi,

It's in Managing MC/Serviceguard manual. You can download it from docs.hp.com

[melvyn burnard]

Please bear in mind, hte serial link hearthbeat is NOT a full heartbeat, and if you have more than one lan over which to send a heartbeat, we do not recommend using hte serial heartbeat. This was origionally desgined for older systems that had a lack of I/O connectivity.
If you do really want to try to use this, hten consult the Managing Serviceguard manual available at
http://docs.hp.com/hpux/ha

[Christopher B. Santos]

Hi,
I've seen the Managing MCSG manual but it does not say how to cnfigure the serial line as heartbeat. My customer is using lan0 as production lan; lan1 used for data backup(legato) and lan 3 as standby lan. They want to have the heartbeat be configured with the backup lan (lan1) and the serial line as well. If my heartbeat (lan1) is not connected by cross cable and being used by data backup as well, there might be network saturation, thus the need for serial line as heartbeat is necessary.

[melvyn burnard · Accepted solution]

What does the ServiceGuard RS-232 heartbeat provide?

The RS-232 heatbeat (also known as the serial heartbeat) is supported in a two-node cluster only. It should only be used if you have a two-node cluster that has ONLY ONE network card available. For example, you may have a low-end D-Class server which has a limited number of slots for network cards. There may be only one card, such as a FDDI or 100BT card, available to be used for a ServiceGuard heartbeat LAN. Under such circumstances, it would be appropriate to use the RS-232 heartbeat. However, if you have at least two heartbeat LANs, or one heartbeat LAN and one standby LAN, a serial heartbeat SHOULD NOT BE USED.
This is because RS-232 lines are inherently unreliable compared to network cards which run the TCP/IP protocol. Unlike TCP/IP, the serial line protocol has no error correction or retry mechanism.

RS-232 lines can also be complex and difficult to configure because of a lack of standards (e.g., 9-pin vs. 25-pin cables, different serial ports support different pin-outs, baud rates, etc.).

In summary, the RS-232 heartbeat should only be used WHEN ABSOLUTELY NECESSARY.

The RS-232 link is only used for cluster heartbeats, and does not support other cluster communication. If all heartbeat LANs fail, one of the two nodes will be removed from the cluster. One of the real values that the RS-232 heartbeat provides is to handle LAN failures more intelligently. For example, in a cluster with a single LAN based Heartbeat, if the network adapter card in node 1 failed, both nodes will detect a network failure. There is an equal chance for either node to win the cluster lock and remain in the cluster, while the remaining node will TOC. So, the odds are 50 percent that the node with the failed network adapter card will be the node that remains in the cluster. This is probably not the desired result.

If the RS-232 heartbeat has been configured in the same situation as described above, the behavior will be different from above. The node with the failed network adapter card will still be receiving the RS-232 heartbeat, so it will increase the delay before it tries to acquire the cluster lock, then it stops sending the RS-232 heartbeat and finally TOCs itself. That will allow the remaining node with a healthy network adapter card to win the cluster lock and to successfully reform the cluster.


NOTE 1
Some releases of ServiceGuard do not have the capability for the user to monitor the status of the RS-232 heartbeat link. Furthermore, no messages of any kind are logged in the syslog for any RS-232 heartbeat link. The only way to tell if the RS-232 link is working properly is to halt the cluster, and perform a manual read/write test of the connection using command line utilities such as cat(1M), or to connect a serial link analyzer between the two ports.


--------------------------------------------------------------------------------

Common problems and known issues with the RS-232 heartbeat

When trying to determine what device file to use for the serial heartbeat, people commonly use SAM to tell them what ports are available. In doing this, it is possible to enable a getty(1M) process for the port, and this getty process will interfere with ServiceGuard and prevent the serial heartbeat link from being established. To determine if this is the case, scan for the serial port device file in the /etc/inittab file, and if it exists comment it out (or remove it), then run "init q" to force init to re-read the /etc/inittab file.

ServiceGuard opens the serial port at 19200 baud; however, some ports do not support this speed. If this is the case, then that port cannot be used as a serial heartbeat link for ServiceGuard. This problem is most commonly found on Multifunction I/O boards (also known as the Core I/O adapter). Consult the documentation for your serial port to determine if the port supports operation at 19200 baud.


--------------------------------------------------------------------------------

Cabling guidelines for the ServiceGuard RS-232 heartbeat
See Chapter 6 of the HP9000 Enterprise Servers Configuration Guide for cabling guidelines. The minimum requirement is that the Transmit, Receive and Signal Ground must be connected to each other in the following manner:


Transmit -> Receive
Receive -> Transmit
Signal Ground -> Signal Ground

The HP9000 Enterprise Servers Configuration Guide implies that these pins are always 2 (Transmit), 3 (Receive) and 7 (Signal Ground). Although this is common, it is not always true. You need to verify the correct pin-outs for Transmit, Receive, and Signal Ground for the cable you are using.
In addition to the cables listed in chapter 6, the following cables have been shown to work for certain D and K-Class servers. Of course, you will need to test these cables with your particular configuration to insure that they are compatible.

25-pin Male to 25-pin Male, HP P/N 92219Q

25-pin Male to 25-pin Male, HP P/N 28659-63005

9-pin female to 9-pin female, HP P/N 5182-4794

[Christopher B. Santos]

Hi Melvyn,
Thanks a lot for the well explained information. I'll try to look in to this and convince my customer not to use serial as heartbeat. Thanks again. :)