CODEWITHSUNDEEP
cassandra
asp
asp

Reputation: 835

Restart sequence for multi node data center

We are using Cassandra version : 3.11.2 and We have two data centers A and B both have 3 nodes each , while seeds i have mentioned as 1st ip of DC A and 1 ip of DC B in all 6 nodes in cassandra.yml.

Now for maintenance for patching on DC B nodes, where we need to stop cassandra on DC B nodes and restart it..

My confusion is:

Restart doubts:

I am asking these question in light of scenario node failure or data loss..

Please suggest.

Upvotes: 1

Views: 103

Answers (1)

Alex Ott
Alex Ott

Reputation: 87144

If you have hinted_handoff_enabled enabled (by default) and your maintenance is inside max_hint_window_in_ms time frame (3 hours), then after restart of DC B the data will be streamed from nodes in DC A, so eventually you get all data into DC B as well. To make sure, you can perform repair after all nodes are back.

Regarding restart:

  • It's better to start seed node in DC B first, but you need to make sure that seed node in DC A is online at that moment;
  • You don't need to stop all nodes of cluster before restart of one DC;
  • in DC B you can even patch your nodes one by one, and restart them separately, but this may depend on what kind of patching you're doing, like, is it Cassandra upgrade, or something like. In this case you may need to disable repairs, etc.

Upvotes: 1

Related Questions