Introduce force flag for new deletion API (#12588)

## Problem

The force deletion API should behave like the graceful deletion API - it
needs to support cancellation, persistence, and be non-blocking.

## Summary of Changes

- Added a `force` flag to the `NodeStartDelete` command.
- Passed the `force` flag through the `start_node_delete` handler in the
storage controller.
- Handled the `force` flag in the `delete_node` function.
- Set the tombstone after removing the node from memory.
- Minor cleanup, like adding a `get_error_on_cancel` closure.

---------

Co-authored-by: Aleksandr Sarantsev <aleksandr.sarantsev@databricks.com>
This commit is contained in:
Aleksandr Sarantsev
2025-07-17 15:51:31 +04:00
committed by GitHub
parent 5dd24c7ad8
commit 4559ba79b6
5 changed files with 111 additions and 50 deletions

View File

@@ -76,6 +76,12 @@ enum Command {
NodeStartDelete {
#[arg(long)]
node_id: NodeId,
/// When `force` is true, skip waiting for shards to prewarm during migration.
/// This can significantly speed up node deletion since prewarming all shards
/// can take considerable time, but may result in slower initial access to
/// migrated shards until they warm up naturally.
#[arg(long)]
force: bool,
},
/// Cancel deletion of the specified pageserver and wait for `timeout`
/// for the operation to be canceled. May be retried.
@@ -952,13 +958,14 @@ async fn main() -> anyhow::Result<()> {
.dispatch::<(), ()>(Method::DELETE, format!("control/v1/node/{node_id}"), None)
.await?;
}
Command::NodeStartDelete { node_id } => {
Command::NodeStartDelete { node_id, force } => {
let query = if force {
format!("control/v1/node/{node_id}/delete?force=true")
} else {
format!("control/v1/node/{node_id}/delete")
};
storcon_client
.dispatch::<(), ()>(
Method::PUT,
format!("control/v1/node/{node_id}/delete"),
None,
)
.dispatch::<(), ()>(Method::PUT, query, None)
.await?;
println!("Delete started for {node_id}");
}