Question about changing expected_votes value

[Cactus]

I want to create a two-node cluster with one quorum disk. From the doc, I konw that the expected_votes should be 3.
But on one node, this value is 1 currently (using pipe mcr sysgen show/all | search sys$pipe "expected"). I want to change the value to 3. But when I issue the command set cluster/expected_votes=3. The expected_notes value I got from sysgen is still 1. Also, from the doc, I found that I can use SYS$UPDATE to change the system parameter, but I can't find EXPECTED_VOTES in the the file MODPARAMS.dat...
Anyone could give me some suggestoins, thanks in advance.

[Hein van den Heuvel]

This is all pretty well described in the various OpenVMS Manuals, FAQs, Wizard and such. Google: +expected_votes +cluster +openvms gives lots of links.

With a quorum disk and two nodes I would expect you wanted expected_votes to be 2.
Any one node + quorum disk should be allowed to get the cluster going.
With expected_votes = 3 you tell the cluster to wait until both nodes and the quorum disk can see each other. Is that what you want?

Cheers,
Hein.

[Dean McGorrill]

Hi Catus,
expected votes is not dynamic. I usually
use sysgen on each node

sysgen> use current
sysgen> set expect 3
sysgen> write current

set disk_quorum to your disk, set qdskvotes 1, votes 1. I usually reduce recnxinterval
if I'm on a quiet lan. hope this helps -Dean

[Dean McGorrill]

oh and add this stuff in modparams.dat

[Andy Bustamante]

Welcome to the VMS forum.

>>> I want to create a two-node cluster with one quorum disk

For a new cluster, you need to have quorum to create quorum.dat on the quorum disk. This can be a chicken or the egg problem. You'll need either use an an interactive boot or sysman and reboot to, first inflate the votes on a single node forcing quorum. Then, boot that node. If the quorum disk isn't the system disk you'll need to mount it. The node will create quorum.dat and the quorum disk will start providing a vote. Restore votes to normal. The quorum disk is usually mounted in SYLOGICALS.COM if it's not the system disk.

As far as EXPECTED_VOTES missing from modparams, feel free to add it. You're correct that it would be set to 3. Quorum will be 2. Did you use cluster_config.com? Make sure it's set properly on both nodes. I'm assuming both nodes will have 1 vote and the quorum disk will also have 1 vote.



Andy

[Cactus]

Thanks for so much quik response. I have another question, shall I define some logical values like CLU$COMMON before I create the cluster on one node, since when I run the cluster_config_lan, there is warning about this. Thanks

[Hoff]

What is the quorum disk interconnection? Multihost SCSI? DSSI? (Where I'm going: If it's not connected on a shared interconnect with at least two hosts, don't bother with a quorum disk.)

One of many discussions of the topic...
http://64.223.189.234/node/153

[Volker Halle]

Hein,


With a quorum disk and two nodes I would expect you wanted expected_votes to be 2.


I'm afraid you've got that wrong. You were probably thinking of QUORUM instead of EXPECTED_VOTES !

For a 2-node cluster and a quorum disk and for one votes for each, you would want EXPECTED_VOTES=3, which will give a quorum of 2. One node + quorum disk or both nodes without the quorum disk will allow the cluster to run.

Volker.

[Hein van den Heuvel]

Volker (and Andy) yeah, I got that wrong I wanted to make sure Cactus coudl run with 2 out of 3 but that mean quorum=2, expected=3.

Volker (and Jon) sorry an other topic ( 1139459 ) I did not check the ppf program you already pointed to an provided my own.
Much the same. My code uses sys$open to get the IFI for the PPF file. Cleaner than setting the PPF bits in the IFI obtained from the logical, but the same effect.

Hein.

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=1139459

[Cactus]

I have another question, shall I define some logical values like CLU$COMMON before I define the cluster, since there is warning about this when I issue cluster_config_lan. Thanks.

[Jon Pinkley]

Cactus>>>"shall I define some logical values like CLU$COMMON before I define the cluster, since there is warning about this when I issue cluster_config_lan."

I think the only time you get that message is when configuring with multiple system disks. I think that is why Hoff asked: "What is the quorum disk interconnection? Multihost SCSI? DSSI? (Where I'm going: If it's not connected on a shared interconnect with at least two hosts, don't bother with a quorum disk.)"

You may wonder how multiple system disks and the question about the quorum disk interconnection are related.

The only time a quorum disk makes sense is when more than one node has a direct connection to it. And if you have a device that is directly connected to more than one system, then it is far simpler to use that device (or another device with direct connections to your "boot" nodes) as a shared system disk with multiple system roots. In a shared system disk configuration you already have a system defined location that can be used as a cluster common area; specifically the sys$common tree on the system device.

There are reasons to use a two distinct system disks in a two node cluster, for example a mixed architecture or mixed version cluster, or just for the educational value; but unless you have a good reason for multiple system disks, I would recommend using a single system disk in the cluster.

Can you tell us a bit more about the hardware configuration of your two node cluster?

Also, if you haven't read it, check out the reference Hoff left: http://64.223.189.234/node/153