2500 upgrade - cluster

[lapluk_]

Hi

i have a cluster running 7.1R9, how i can upgrade to 7,2.R4 without downtime?is there any how to ?

L

[SHKM_]

When we upgrade the system it needs reboot, so downtime needed.

1. Read the release notes of the version to be upgraded

2. Take a backup

3. Start upgrade

You can refer the admin guide for the backup and upgrade steps. Here is the link to download the admin guide, release notes, latest SA software version.

http://www.juniper.net/support/downloads/?p=sa6500#docs

Thanks,

[lapluk_]

i saw realse notes but there is no info how to upgrade cluster, i think there shoudl be a way to do it without downtime,

how about disabling stanby from cluster, upgrade, enabling cluster and upgrade other one? somebody try that?

[muttbarker_]

Yes - if you are running active / passive then you upgrade your cluster with no downtime. Your steps are good - 

Disable the backup, upgrade it, enable it

Disable the master, upgrade it, enable it

And it works just fine.

[-red-_]

Unless I am missing something about the details of Juniper clustering I don't understand what is gained by disabling the peer?

[Felipe_]

Hi,

Disabling the node or peer in the cluster would ensure that the box wil not accept user connections and that it will not participate in syncing sessions/configs while it is being upgraded. After upgrade is completed and you are fine with users failing over to this newly upgraded node, when enabled, it will start upgrading the other node theoretically and eventually failing over VIP to the first one.

The other option to do is do not disable any node in A/P, but upgrade first the passive node (not serving users), and once it is finished upgrading, it will push the package to the other, and then it wil failover to the first one you upgraded -- hence no downtime.  The users using NC will not lose the session, but next time he logs in, he will receive the new client for that new release.  Sometimes, there may be hickups so I recommend really to do this in a maintenance window.

Regards,

Felipe