-
Notifications
You must be signed in to change notification settings - Fork 488
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
timeline_ancestor_detach: hardlinking where possible #8828
Comments
@mtyazici just ran into this today: he wondered why it takes 50 seconds after an ancestor detach for the compute to be able to start. I would say this issue is responsible. the compute spends those 50 seconds in whatever gets into And indeed, we see on-demand downloads in the 50 seconds after the finishing of the ancestor detach in logical size calculation, which do download previously "remote copied" layers:
|
For all of the small number of uses timeline ancestor detach (#6994) has seen, there is a need to download layers in the restart following completion of the prepare phase after
reset_tenant
. The on-demand downloads happen for logical size calculation or compaction.We could simply hard link all
Layer
instances which happen to be present already.Layer::keep_resident
can be used to test for the residency, and keeping the layer resident while hard-linking happens. There are already TODO comments, and the way we fsync might need to change as well.This is a performance optimization, but it is easy to do.
The text was updated successfully, but these errors were encountered: