upgrading cluster from r80.40 to r81.20 using smartconsole upgrade feature

[Luis_Miguel_Mig]

I am trying to upgrade a cluster from r80.40 to r81.20 using the  blink image r81.20 + latest jumbo.
the file is 8Gb, so the download takes a while.
The process has failed with a timeout error. I was wondering if there a timeout value configured for this download that could be increased. 
Any idea? 

[AkosBakos]

Hi @Luis_Miguel_Mig 

To be honest this feature is not my friend :-). I feel I don't have enough control on the process.

In which part of the upgrade happened the timeout?

What I know exactly the policy install timeout is 10 minutes

BR

Akos

[Luis_Miguel_Mig]

the timeout error failure happened during the firmware download.

[Luis_Miguel_Mig]

I get this error in smartconsole when the download is about 45% but funnily enough cpuse is still downloading and making progress after the smartconsole error. See screenshots attached

[AkosBakos]

Hi @Luis_Miguel_Mig 

323Kb/sec is incredible low speed nowadays. 

I don't know, maybe there is a reason, but why don't you try to do in on the gateway itself?

1. copy the blink package via scp
   
   1. it will take for a while
2. import it (in CLISH)
   1. installer import local
3. Verify it
4. then install is

I usually use this method.

Akos

[the_rock]

Im sure someone else may tell you otherwise, but based on all the testing I had done in the lab with different versions, I would NOT bother with doing this method when it comes to major version upgrades. Minor ones and jumbo fixes are fine. Thats just my honest feedback based on all the lab tests I completed.

For major upgrades, I ALWAYS stick with web UI option and works fine.

Andy

[Luis_Miguel_Mig]

I have finished the cluster upgrade in my lab cluster and it has been a great experience.
smarconsole looks after pretty much everything and it provides good information in real time about what is going on.
These are the steps I followed:

1) I pre downloaded the r81.20 image in both cluster's gateways
2) install on non active member only, untick the failover option

3) manual failover:  clusterXL_admin down; clusterXL_admin up; cphaprob stat

4) install on non active member only, untick the failover option
5) deactivate mvc: cphaconf mvc off

6) install the policy (not really required as smartconsole did it automatically, but good to see that policy install accelaration works and the policy got installed in no time)

Summary:
I  think smartconsole gradual upgrades is a great feature, but I would like to see a couple of improvements in the future releases:
1) Better firmware download management - in sync with cpuse
2) Automatic MVC deactivation at the end of the process
3) More verbose mode to present all the cpuse process status information in real time in smartconsole

[Bob_Zimmerman]

My team uses this feature all the time, and yes, getting the package onto the systems is the most annoying part. I'd recommend telling the members to download it days ahead of time. When it's time to actually do the upgrade, tell CDT to work with the package on the members instead of pushing it from the management.