Data Center (VDX)

Hello,

I'm going to upgrade my cluster of VDX6740T-1G. From documentation, I found that I have to go one major upgrade at time. I have made this question a lot of time and also open a case asking for recommended path and it is not very clear (at least for me).

I'm at 4.1.3 version, so I had choosen this path:
4.1.3 to 5.0.2c with coldboot option
5.0.2c to 6.0.2f with coldboot option
6.0.2f to 7.0.1c with coldboot option
7.0.1c to 7.1.0b with coldboot option
7.1.0b to 7.2.0 with coldboot option
7.2.0 to 7.3.0 with coldboot option

The other thing is that I'm a little confused with the way I have to do the upgrade. Documentation said there's three ways to do it:
1. All devices at once
2. Odd/even approach
3. One device at a time

So, my question here is: if I use option 2 or 3 (as recommended for logical-chassis), is it possible that I loose at some point the remote access to one of the leaf nodes for example?

Lets say, I use the one by one approach and for some reason the VCS is not capable to came UP after some of the versions upgrade. If that happens, i will lost management access (sure I have a computer at the console port, but I'm trying to think in the worst scenario). Is it a way to check compatibility between versions?

Since I have to go through all this versions, what should be the best way to do it? all devices at once for each step? that way I'm assuring they came up with the same version...

Another things, there's a lot of possibilities with the command firmware upgrade... what is the recommended one?

Sorry to post all this questions, I'm newly to brocade and we had already some troubles with these devices. These VDX are used into our DataCenter, so they're very importants and I have to minimize the risk here.

Thanks in advance for your collaboration,

Laura

Laura,

Thanks for your question. First of all, if you are not receiving the support you need in your case, please request to have it escalated or for a Manager review. This will make sure your questions are answered.

Regarding Firmware upgrades, there are three main methods:

• In Service Software Upgrade (ISSU) when going from one minor release to another. This is a non intrusive, warm reload of the VDX and can be completed on an whole fabric without losing traffic.
• Cold Boot, when going between major releases. This requires the VDX to be fully reloaded and causes a traffic interruption
• Default-Config, when making large jumps in major releases. This is a preferable option when you have a smaller config and are looking to save time with an upgrade that would require multiple steps.

Depending on your design the order in which you complete your upgrade can change. Normally with dual homed hosts in Data Centers we recommend using the "Odd Even" approach.

When using this approach, you should not lose management access to the devices, the Management network is independent of the VCS fabric and would be reachable even if all the devices are segmented.

Regarding compatible versions, a VCS fabric will only form if all devices are on the same version, so with this said, when completing the Odd/Even approach we recommend to fully upgrade the Odd side (excluding your Principle switch), then upgrade your Even side (excluding your Principle switch) and then finally upgrading the Principle last.

For the actual Firmware Download command. I would recommend you use the "interactive" option. The VDX will prompt you for all the necessary options.

• firmware download interactive

I hope I covered all of your questions, or at least gave you some clarity. Please also review our KB article for Best Practices when completing a Firmware Upgrade:
https://gtacknowledge.extremenetworks.com/articles/Solution/NOS-Firmware-Upgrade-Best-Practices

And if you need anything further, please do not hesitate to escalate your case and we can discuss further.

Thanks,

Mike Morey
Principal Technical Support Engineer

Laura,
I am the Lead Data Center engineer and am responsible for 3 large data centers that use the VDX platform and have done many upgrades over the years so I have a lot of experience doing what you need to do. We use all flavors of VDX including 8770's, 6940's and 6740's. I would like to make a couple suggestions prior to the upgrades. Make sure each fabric has a shared IP management IP address, make sure each switch has a console connection, make sure the principle priority is set for switches within the fabric so there is predictability as to what switch is the principle switch during the upgrade reboot process. Document all these connections and addresses so that you know where and how to connect if there is an issue.
We use the 3 step approach, having every server connected to 2 switches (odd and even), we upgrade the odd, then the even and lastly the principle switches waiting for each step to complete prior to beginning the next (approx. 45 min for each on a 6740T-1G's). We do the entire data center using these steps. For example if we have 50 fabrics of 8 switches per fabric, we do all fabrics at once so step one would be upgrading 200 switches, then step 2 would be upgrading 150 switches then the final step would be upgrading the final 50 principal switches. We can complete a single upgrade or an entire data center in less than hours by using the approach but it is essential that you verify all switches in the step are upgraded before moving onto the next group. If it were me I would upgrade all fabrics from one version before moving to the next. I hope this helps.

Thanks a lot to all of you.
@John...A BIG thanks for share your experience.

I will take your advice. I don't have any brocade experience (I worked before with Fortigate, HP, Cisco) and it was all the time for SME. It's the first time I'm going to do an upgrade for Brocade VDX and for a DC of this size also.

So, thanks again.

@Michael: thanks for the link and your comments. That link was the first document I had consulted in order to create my upgrade procedure. I have not complaints with extreme support. I had a case open asking for the recommended upgrade version, but they just sent me to the knowledge base (which i already have done before open the case)

I just though that there was a recommended version for each model. For example, I'm also doing the Upgrade for the Brocade ICX7450 I have at access level. When I looked at the information (in the Ruckus web site) they recommend to continuous into the same release and only apply the last patch. If there's some feature you need that is added in higest release, then you can follow the upgrade path and change your release.

If I understood right, for VDX and NetIron (Brocade models bought by Extreme Networks) the approach is different.

Any way, thanks to you both. I will take your advises.

Best Regards,

Laura

Laura,

Historically, Legacy Brocade followed a method called "Target Path". While there are new releases all the time, they would recommend to upgrade to a specific code path unless there were features in a newer code that were needed. This is to ensure that the most stable version is recommended to customers in general.

Extreme has continued this recommendation style (it sounds like Ruckus/Arris has done the same). The official documentation for Extreme Network OS is here:
https://documentation.extremenetworks.com/networkos/HW/extreme-nos-target-path.pdf

I can understand that this document is confusing as it lists multiple versions. In my experience I like to see customers running the latest letter patch of NOS 6.0.2. If you are interested in IP Fabrics, then we would recommend to moving onto NOS 7, otherwise it is fine to remain on the NOS 6 release.

Thanks,

Mike Morey
Principal Technical Support Engineer

There is a little mistake there in document for NOS6.0.2,
Currently the latest and recommended version is 6.0.2h

Thanks Yulia! 🙂

Thanks a lot Michael. In that case, I think will be fine at NOS 6.0.2. This version will be OK for now.

Thanks a lot again, your comments had been very helpfull for me. I hope everythings goes ok with my upgrade... First time are always hard ;)

Thanks again and have a nice day ahead!