X450a-48t IPv6 BGP configuration and version compatibility

Hello, I have a stack of X650-24x and X450a-48t, both with Core licenses.

Both switches are currently running XOS v1265b2-patch1-3, however I need IPv6 BGP capability which is not supported in this XOS version.

What version of XOS can I use to get IPv6 BGP capabilities for both these switches? I have read somewhere that it is not advisable to upgrade X450 to above 12.6, I believe this is only if you have a very large routing table (25,000+ routes)?

I currently only have a small routing table and will never have anywhere near 25,000 routes. Will I still have problems if I upgrade my stack to 15.1.4?


9 replies

Userlevel 4
I would reccommend you to upgrade code to patch.
then try the configuration.
Userlevel 2

IPv6 BGP capabilities has been added from EXOS 15.1, so you need to upgrade at least in this release.
Summit x450a is fully supported up to 15.3, included. The only "challenge" here, especially in the case of BGP, is the memory usage. Summit x450a has less memory than more recent hardware platforms, and as new major version of EXOS are released, the memory footprint is bigger. As BGP often means large RIB, this might be a problem in certain cases.

Just validate your memory usage when you upgrade, to see if that can be a problem for your case. You may want to monitor it as well, for the long run.
Thank you for the clarification, so I should be ok to install on both switches?
Userlevel 4
Use latest patch . Stable version .
Thanks. I don't have support so have to live with what I can find online. I have found so that should be fine, is the latest of that version.
I have just checked memory usage:

on the X450:
Mem: 230276K used, 24380K free, 0K shrd, 34764K buff, 36960K cached

on the X650:
Mem: 588716K used, 427728K free, 0K shrd, 131780K buff, 263400K cached

The X650 is the stack master, are the routing tables stored in the memory of both switches?

The memory usage on the X450 seems to be taken up primarily by 1 main process:
1428 1 root S 29272 11.4 0 0.0 ./cliMaster
1424 1 root S < 9276 3.6 0 2.6 ./hal
1570 1 root S 5136 2.0 0 0.0 ./etmon
1585 1 root S 3212 1.2 0 0.0 ./xmld
1418 1 root S 2552 1.0 0 0.0 ./emsServer

The cliMaster process is taking up 11.4% of memory, i'm guessing this is an important cli process and there is no way to reduce the memory usage of this process?
Userlevel 2
I forgot you had a stack with x650. I believe RIB is only on the Master node, as it's acting as Control Plane for the stack. I don't think the Backup has a copy of it (I'm not 100% sure here).

What's your stack made of? 2 units, one x450a and one x650?

Reading your output, I'm wondering if you would have enough memory on the x450a in the event of a x650 failure.
It's one x650 and one x450. I would guess the rib has to be stored on the backup too incase of the master failure? I'm not really using the x450 as a backup, they are just in a stack for convenience.
Userlevel 2
You should look at memory usage of the BGP process, as well as your IGP, to see if the x450a would have enough memory. Obviously, if your BGP peers are only on the x650, there's no issue as a failover would more likely result in losing them at the same time.