how to migrate one node to a new node

[pccwglobalit]

we have four nodes NLS. However, due to lack of disk spaces and we need migrate all data to another log server. please advice the steps to migrate. thanks.

[jolson]

pccwglobalit,

The best way to approach this is to spin up your new server and join it to your existing cluster.

Now, one node at a time, shut down the rest of the cluster. Ensure that your cluster is in a green state in between node shutdowns. Eventually all of your data will migrate to the new node.

This whole process will take time . Be certain that your cluster is in a green state before shutting down further nodes.

[pccwglobalit]

thanks.
we only purchase 4 node licenses and can we add one more node for the time being and after finished migrated, we then remove unwanted node. is that ok?

[jolson]

That shouldn't be a problem. You may get a license complaint, but you should be able to access the Administrative page as necessary. After the new nodes is attached and the status is green, shut one of your old nodes down and you should be good to go from there!
.

[pccwglobalit]

due to one node limited resources, now the cluster is relocating some shards to that node so the cluster status is red. Could we just setup another node and then shutdown the limited resource node?

[jolson]

It's hard to say at this point. You should be able to shut down the limited resource node, but there is no guarantee that data loss will not occur, since the cluster status is already red. Is there any way to temporarily expand the disk of your failing node?

[pccwglobalit]

i have removed and closed some index. however, there are still some shards relocating to that limited resources node. The problem is that the link is too slow to relocate.
i want to know how many shards need to relocate to that node.

[jolson]

That is very hard to determine without seeing your environment. If you are okay with sending a ticket in, I would be happy to assist you remotely. If you have backups in place and working, you should be able to remove the bad indices in question and restore from your backups - would that work for you?