Frozen tier - Conenience in multiple zones

[Alberallo]

HI,

since high availability is guaranteed by default for frozen nodes, why should there be any convenience in configuring a cluster with more than one zone for the frozen tier?

In particular, during the execution of the ILM policies, how does Elastic act if a frozen policy is set but a node of that type is not available?

Thank you,
Alberto.

[warkolm (Mark Walkom)]

Welcome to our community!

Just to be clear, are you referring to Data tiers | Elasticsearch Guide [8.7] | Elastic?

[Alberallo]

Hi @warkolm,

thank you for your reply.
Yes, I'm referring to Data tiers topic and in particular on version of Elasticsearch is 7.17.6, but I think that this concepts are general for the product.

I'm available for any other question and I hope you help me to understand how cluster work if it is configured a single frozen node and it's necessary to read and especially to write (during execution of ILM policy) data from it but it is unvailable.

Thank you so much.

Regards.
Alberto.

[Christian_Dahlqvist]

@Alberallo

since high availability is guaranteed by default for frozen nodes, why should there be any convenience in configuring a cluster with more than one zone for the frozen tier?

I am assuming you are referring to the seacrchable snapshots feature backed by S3 storage. I do not think you need to configure zones (although it probably does not hurt either) as you hold at most the primary shard, which is backed by S3 data. You should however make sure you distribute your nodes across multiple availability zones so you do not lose all nodes if an availability zone goes down.

[Alberallo]

@Christian_Dahlqvist

I am assuming you are referring to the seacrchable snapshots feature backed by S3 storage. I do not think you need to configure zones (although it probably does not hurt either) as you hold at most the primary shard, which is backed by S3 data. You should however make sure you distribute your nodes across multiple availability zones so you do not lose all nodes if an availability zone goes down.

Hi @Christian_Dahlqvist,
thank you for your reply.
It's correct what you suppose: I'm referring to searchable snapshot. It's logical that with multiple nodes in multiple zones I increase my reliability but as explained in my last reply to @warkolm I want to understand in deep what happen if I have one frozen node, it goes down, and for example the ILM policy need to move data into this node. About read I assume that during the unvailabilty period the data will not be searchable, but when node will be available the data can be searched again.

Thank you.

Alberto.

[Christian_Dahlqvist]

If you only have 1 frozen node and ILM should move data to this and it is not available the data will as far as I know stay on the node where it is. I would always recommend having multiple nodes per tier.

[Alberallo]

@Christian_Dahlqvist

If you only have 1 frozen node and ILM should move data to this and it is not available the data will as far as I know stay on the node where it is. I would always recommend having multiple nodes per tier.

Hi @Christian_Dahlqvist,
thank you for your reply. Do you know if there is a SLA that Elastic must gurantee in the recovery of the node down?

Thank you.
Alberto.

[Christian_Dahlqvist]

I do not know. Is this cluster hosted on Elastic Cloud? If so I would reach out to Elastic Cloud support about that.

[Alberallo]

@Christian_Dahlqvist

I do not know. Is this cluster hosted on Elastic Cloud? If so I would reach out to Elastic Cloud support about that.

Yes, the cluster is hosted on Elastic Cloud.

Thank you.

Alberto.