SA 6500 cluster (Active/passive)

[mzhaim_]

Hi i have created an active/passive cluster of two SA 6500 devices and everything is working fine except i would like to know how can i change the cluster leader from one device to another also how can i ensure that this device is always the leader (if it goes offline and comes back again it should be the leader when it is online)

thank you for your help

[Jickfoo_]

Supposingly it is the higher sync rate that determines who the leader is. Although, I recently had a failure and my cluster never failed back to the leader. Perhaps its a bug, I am running 6.0R7.

If you want to manually fail over, under clustering status select the primary and choose disable, it should fail to the backup unit.

Best of luck,

Justin

[haas_]

I have not had good success with it rolling back either on a pair of 6500's. I have had to manually do it. I am now on 6.1R5 and haven't tested it yet on this code.

[Mel_]

I think you need to perform a restart of the cluster services in order for it to take any notice of the sync rank. It's not very well documented though!!

[Mrkool_]

well i have 1 pair of sa2000 and 3 pairs of sa6000s and none of them revert back to primary i always have to do a manual failover. The rank only kicks in if you do a restart of the cluster or the cluster services as Mel mentioned.

[klind_]

In working with some SA4500's it seems there definately is a bug for failing back to the primary device. Sometimes it works, other times it doesn't. One thing to note about the SA4500's is you should not need to restart the cluster. If you change the sync rank and click the update button, the devices "should" sync and make the appropriate device the leader. Just as an FYI here is the description of the sync rank column in the clustering status tab provided by Juniper:

Specifies the synchronization order for nodes when rejoining a

cluster. Accepts sync ranks from 0 (lowest rank) to 255 (highest

rank). The highest rank takes precedence. Where two nodes have

identical sync ranks, the alpha-numeric rank of the member name is

used to determine precedence.

[ruc_]

Note: In an A/P cluster the 'leader' node does not guarantee that this node will own the VIP (i.e. handle user requests). A leader node simply means that when nodes rejoin (after a cluster split) and in the rare event that the clustering algorithm is not able to decide which node has the latest config then the leader nodeÕs config will be considered as the latest config and will be used to overwrite the config of other nodes.