bl860c Itanium Blades in enclosure with Flex10 VC modules.

[The Brit]

This is an informational message for anyone who is running or is considering running bl8x0c blades in an enclosure with Flex10 VC modules (note this might not apply to the new "i2" blades)

This issue has been mentioned and/or discussed in several earlier threads, however I wanted to put out a solution which appears to get around the problem.

The PROBLEM.

The onboard network devices on the bl8x0c blades, running OpenVMS, default to AUTONEGOTIATE=enabled.
The downlinks on the Flex10 VC module will not negotiate.
HP recommends that the Flex10 downlink ports be set to "1Gb Custom" setting, and that AUTO-NEGOTIATE be disabled at the VMS (LANCP) level.

Scenario:

You have an existing Blade Cluster, and you want to add a new blade as a new cluster member.
1. Run CLUSTER_CONFIG
2. Try to boot the new blade into the cluster.

Result:

During initialization, network devices are intitialized, (right after "Waiting to join or form a VMS Cluster"), however the links do not come up (because of the "autonegotiate" thing) and therefore your membership request never reaches the existing cluster. Startup will hang at this point. (Note also that since this is the first boot into the cluster, votes are set to 0, to protect the system disk, and so the hang never ends)

It is possible to get past this by;

1. shut down the cluster.
2. conversational boot of new blade.
3. set votes=1 expected_votes=1, continue
4. blade will boot up as first node in cluster, allowing you to log in and configure the network devices (i.e. disable autonegotiate.)

It is then necessary to shutdown the node, reboot the old cluster, and then reboot the new blade.

Obviously this is a very cumbersome procedure.
--------------------------------------

Thanks to some discussions with Colin Butcher, it appears that there is a simple shortcut which solves the problem. It might even be optimal given that the bl8x0c blades have pretty much identical hardware.

1. Run cluster_config as normal
2. from the existing (blade) node, copy SYS$SPECIFIC:[SYSEXE]LAN$DEVICE_DATABASE.DAT (and .SEQ), to SYS$SYSDEVICE:[.SYSEXE]
3. Do conversational boot of new blade.
4. Check that VOTES=0, EXPECTED_VOTES=<# of voting systems currently up>
5. Continue.

You should see (after the "Waiting to form or join a VMSCLuster") the Network Devices, not only initialize, but you should also see the links come up. This all occurs before the node joins the cluster.
The node should then continue to join the cluster run net$configure and autogen, and then reboot, as you would expect.

Again, thanks to Colin for the suggestion, and I hope this helps someone in the future. I guess that many in the forum would have come up with this solution independently, however for me to figure these things out, I need it to be engraved on the head of a 3lb block hammer and then have someone crack me over the skull with it (the light goes off when I regain conciousness).

Disclaimer: I dont know how applicable this solution would be, and I wouldnt recommend it, when there is a significant difference in the hardware between nodes.

for blades however, where the hardware configuration is more or less pre-ordained, it does seem like an acceptable option.

Additional benefit, If your existing cluster node has (locally) standard configuration options, i.e. lla devices for NIC teaming, etc, then this is also automatically set up on the new blade.

Dave.

[The Brit]

This was for information only.