Operating System - OpenVMS
cancel
Showing results for 
Search instead for 
Did you mean: 

OpenVMS Clustering 2x BL860c's - NIC config question(s)

 
SOLVED
Go to solution
Peter Frauenknecht
Occasional Visitor

OpenVMS Clustering 2x BL860c's - NIC config question(s)

Hi everyone.

We'll soon have two bl860c's, to be put into different c7000 enclosures, and to be clustered.
The blade comes with 4 nic's.
-> Two of them will be dedicated to SCS traffic.
-> Two of them are available for the production LAN. How to configure them on the OS level, to get the best results for TCPIP,DECNET,LAT ?
Failsave IP is nice, but could a LAN failover setup also work in a VC environment ?

Please share your thoughts !

Thanks in advance.
8 REPLIES 8
John Gillings
Honored Contributor

Re: OpenVMS Clustering 2x BL860c's - NIC config question(s)

Peter,

Very open ended questions - "best results" depends on your objectives. Performance? Reliability? Also on your loads. Are you really going to use LAT? (Why?) How much DECnet vs TCPIP traffic? How many addresses will you use?

Is this 100MB or Gigabit?

LAN failover works for all protocols, but only has a single device active at one time.

Failsafe only works for IP, but can load share across devices.

What do you mean by "VC environment"?

Why do you believe you need two dedicated SCS adapters? Are you using MSCP to share cluster storage? I'd have expected SAN, so very little MSCP traffic. What traffic are you expecting?

The most important thing you can do, regardless of protocol, load, objectives or anything else is to make sure that EVERY PORT is set to AUTONEGOTIATE.
A crucible of informative mistakes
The Brit
Honored Contributor
Solution

Re: OpenVMS Clustering 2x BL860c's - NIC config question(s)

Peter,
Your 4 onboard NICs will appear to OpenVMS as EWA0, EWB0, EWC0 and EWD0. EWA and EWC are hard mapped to the Enclosure IC Bay 1, EWB and EWD are hard mapped to Bay 2.

Assuming that you have VC Ethernet modules in Bays 1 and 2, then a failover set LLA0, which uses EWA0 and EWB0 will give you the equivalent of NIC Teaming. This configuration works great in VC environment. This is actually very important when you are doing VC Firmware upgrades since the upgrade involves cycling the VC Modules. This will cause the the LLA device to switch physical devices virtually instantaneously, where as the VCM can take upto 35-45 seconds to complete failover.

You could create a second failover set using EWC0 and EWD0 and use this for SCS if you wish. This is not necessary however since SCS will use any and all paths available (including EWC or EWD, depending on which VC module is Active) and will automatically failover when the VC Modules failover.

Note, if your two enclosures are going to be in the same rack, you could create a single VC Domain by adding 10Gb (CX4)stacking links between the enclosures. This would mean that your SCS traffic never has to go to the external network.

Dave.
Peter Frauenknecht
Occasional Visitor

Re: OpenVMS Clustering 2x BL860c's - NIC config question(s)


@Dave: Thank you very much for your suggestions. A LAN failover config it will be then (as I'll be able to assign ONE logical IP and DECnet address to the blade). Any gotcha's to be aware of ?

Robert Atkinson
Respected Contributor

Re: OpenVMS Clustering 2x BL860c's - NIC config question(s)

Peter, I currently have 4 clustered BL860's doing just this.

We have 2 Cisco switches on the private SCS side to give redundancy, one connection per server to each switch.

We force IP and LAT down the public side, then DECnet and SCS down the private side (don't need any external DECnet).

Here's some bits that will help :-

$ DEFINE /SYSTEM /EXEC LAT$DEVICE "EWD0:" ! Force LAT to use this device

$ DEFINE /SYSTEM DNS$ETHERNET_DEVICE " _EWB0: _EWD0: " ! Disable DECdns on these NIC's
$ DEFINE /SYSTEM /EXEC LAT$LINK "EWD0:" ! Force LAT to use this device

$ MC LAVC$STOP_BUS EWD
$ MC LAVC$STOP_BUS EWB



DECnet is done through NET$CONFIGURE :-

"Stop DECnet traffic by configuring :-
@SYS$MANAGER:NET$CONFIGURE ADVANCED
Option 3 â Configure devices on this system
You will be presented for options for each interface enter CSMACD or NONE for that interface"

Hope this helps.

Rob.

Ian Miller.
Honored Contributor

Re: OpenVMS Clustering 2x BL860c's - NIC config question(s)

If you make a failover set called LLA for EWA, EWB then configure IP and DECnet to use LLA

Use SCACP to stop SCS traffic on LLA and EWA, EWB - this will leave EWC, EWD for cluster traffic.

and do

$ DEFINE /SYSTEM /EXEC LAT$DEVICE "LLA0:"
$ DEFINE /SYSTEM /EXEC LAT$LINK "LLA0:"
____________________
Purely Personal Opinion
John Gillings
Honored Contributor

Re: OpenVMS Clustering 2x BL860c's - NIC config question(s)

LAVC$STOP_BUS is a very big and very old and very clumsy hammer (or rather, axe). Much better to use SCACP to set priorities on interfaces than just chop them off.
A crucible of informative mistakes
Robert Atkinson
Respected Contributor

Re: OpenVMS Clustering 2x BL860c's - NIC config question(s)

I guess this is the way "we've always done it", but it does have the advantage of being instantly visible if we loose the hardware for any reason.

Rob.
Peter Frauenknecht
Occasional Visitor

Re: OpenVMS Clustering 2x BL860c's - NIC config question(s)

Thank you !
Your input was really helpful, and is very much appreciated. Thread closed, points awarded.