Saturday, January 18, 2014

How to shutdown a whole Clustered Data ONTAP? I believe I have got the answer

Entrance of Pear Coffee Shop (梨子咖啡館,中科店),
in Xitun District, Taichung City, Taiwan.

A DCADM student several months ago raised a very interesting question: "How to shutdown the whole Clustered Data ONTAP?"

I believe now I have got the answer. It is by executing this command and then power off all nodes.
system node halt -node * -skip-lif-migration true -inhibit-takeover true
I verified the shutdown was clean on a two-node cluster running Data ONTAP version 8.2. I believe it would also work for more nodes.

I cannot answer it right away when the question was raised. After some studying, I found this question is not as trivial as I expected. I just could not find any official document mentioning clearly about how to shutdown the whole cluster.

I found this page at first. Data ONTAP 8.2 Reference - system node halt

However, it only demonstrates how to shutdown one node. And at that time I did not notice the awesome power of "-node *", neither for any additional operand switches. I was also quite concerned that "halt" command might corrupt some RDBs.

I can now tell why the command is the solution.

Switch "-inhibit-takeover true"

This would stop aggregates from being taken over. So, VLDB (one of 4 RDBs) would not be changed at all.

Switch "-skip-lif-migration true"

This would stop LIFs failover to other physical network ports. So, VifMgr (another one of 4 RDBs) would not be changed at all.

For the remaining two RDBs, that is, MGMT and BCOM, they would not be modified at all if we stopped to execute other configuration tasks.

So after we execute above "system node halt -node *", all RDBs would be un-modified and all nodes would be shut down at the same time.

Now I understand RDBs would never be corrupted this way! I hope it helps!

No comments:

Post a Comment