Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[CCR] Define a good default for the soft delete retention policy #34908

Closed
martijnvg opened this issue Oct 26, 2018 · 6 comments
Closed

[CCR] Define a good default for the soft delete retention policy #34908

martijnvg opened this issue Oct 26, 2018 · 6 comments
Assignees
Labels
:Distributed Indexing/CCR Issues around the Cross Cluster State Replication features v6.5.0

Comments

@martijnvg
Copy link
Member

Currently the default for index.soft_deletes.retention.operations setting is zero. Before we release 6.5.0 we should come up with a reasonable default. Coming up with a good default is tricky, because the index.soft_deletes.retention.operations setting is operation based, but we should settle on a number that allows for soft deleted documents to stay around long enough for CCR while at the same time keeping in mind that keeping soft deleted documents do consume resources.

@martijnvg martijnvg added blocker :Distributed Indexing/CCR Issues around the Cross Cluster State Replication features v6.5.0 labels Oct 26, 2018
@elasticmachine
Copy link
Collaborator

Pinging @elastic/es-distributed

@bleskes
Copy link
Contributor

bleskes commented Oct 26, 2018

We can also have non straight forward options for the default - things like 25% of the current doc count with a maximum of 1M (numbers are just illustrative).

@dnhatn
Copy link
Member

dnhatn commented Oct 27, 2018

I opened #34943 for a time-based policy. I will work on a ratio-based policy.

@dnhatn
Copy link
Member

dnhatn commented Oct 27, 2018

How about implementing a size-based retention policy? We can translate the configured size to a number of operations by estimating the average size of operations. It's not entirely accurate but good enough. @bleskes WDYT?

And suppose we have both age-based and size-based policies, the final decision should be ANDed (like the translog policy) or ORed?

@dnhatn
Copy link
Member

dnhatn commented Nov 5, 2018

Using 0 as the default retention is also not good for peer-recovery with soft-deletes enabled. This is equivalent to disable translog retention.

@dnhatn
Copy link
Member

dnhatn commented Jan 31, 2019

Closes in favor of #37165.

@dnhatn dnhatn closed this as completed Jan 31, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
:Distributed Indexing/CCR Issues around the Cross Cluster State Replication features v6.5.0
Projects
None yet
Development

No branches or pull requests

4 participants