Cluster CU didn't success

[Dawei_Ye]

Hi guys,

I am upgrading a R77.20 Cluster to R77.30, but connections had been initialled found after we ran "cphacu start" .

so the CU failed,and after I check the CU Best Practice,I missed:

b) In the Install Policy window, go to Installation Mode > Install on each selected gateway
independently section and clear For Gateway Clusters install on all the members, if it
fails do not install at all.

I didn't uncheck the option,but I installed policy to Cluster success.As I expected ,it would failed due to the old version GW.But it shows all succeed ,is this right?or this step lead the connections table corrupted or something else ?

Any idea with this issue?

Thank you a lot .

Regards,

Dawei Ye

[AlekseiShelepov]

I did an upgrade from R77 to R77.30 today myself, and policy installed without any issues on both cluster members too (R77 and R77.30). I think this is due to R77.X versions being pretty similar, although it wouldn't work like that we previous versions, as I remember. For example, you can manage R77.30 gateways from R77.20 management server.

If you had the setting for policy installation checked and it installed successfully, then nothing went wrong on that step. You would just get a message that policy installation failed on both cluster members, as it failed to be installed on one of them. So then you would disable the parameter and try to install policy again. This is not a problem.

I didn't really used cphacu commands during my upgrade, I just checked that status of upgraded node is Ready, checked connections number with fw tab -t connections -s and that was it. But I didn't have very strict requirements.

I suspect that some sessions were not synched in your case because of enabled SecureXL.

I believe it was not a VSX setup in your case, was it? 

[Dawei_Ye]

Hi Aleksei,

Thanks for your reply.

yes,it was not a VSX environment,just a normal setup.

We turn off the SecureXL before we upgraded the Active member after we run "cphacu" commands.At this point, the upgraded one status is Ready.

But now we doubt is because that we need to uninstall the hotfix installed in GWs before.and the former Active member will prompt the HA and CU status to Active.Then this cause the connections lost.

This time we meet a not really strict requirements, but this customer need to do a reall zero-downtime upgrade next time.So I need to find out where went wrong.

[G_W_Albrecht]

CU Best Practice is good, but i also can recommend Gaia Installation and Upgrade Guide R77 p.138ff, Upgrading ClusterXL Deployments.