feat(1-1-restore): set tombstone_gc mode to 'repair' for views

[VAveryanov8]

This includes views into alterTGC stage as view is a normal table under the hood.

Refs: #4261

P.S. For now it's not clear what the behavior of Secondary Indexes:

1. we don't know if the tombstone_gc mode is inherited from base table. DESC statement doesn't show options for indexes, see Describe schema with internals is missing Index options scylladb#22849
2. It looks like there is no way to provide tombstone_gc mode during index creation or any other properties Add syntax for create a secondary index with synchronous view updates scylladb#16454

---

Please make sure that:

• Code is split to commits that address a single change
• Commit messages are informative
• Commit titles have module prefix
• Commit titles have issue nr. suffix

[Michal-Leszczynski]

Let's also wait for a response to asked question.

[Copilot]

Pull Request Overview

This PR updates the tombstone_gc mode functionality by setting it to "repair" for views along with tables. Key changes include refactoring the view retrieval logic in worker_views.go to use a schema lookup, updating tombstone_gc mode functions in worker_tgc.go to handle both tables and views, and enhancing the integration tests to validate these changes.

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
pkg/service/one2onerestore/worker_views.go	Refactored view retrieval to use a described schema map and removed dummy metadata creation
pkg/service/one2onerestore/worker_tgc.go	Introduced unified functions for getting and setting tombstone_gc mode that handle both tables and views
pkg/service/one2onerestore/service_integration_test.go	Updated integration tests to validate tombstone_gc mode for both materialized views and secondary indexes (with secondary index check commented)

Comments suppressed due to low confidence (2)

pkg/service/one2onerestore/worker_views.go:94

• [nitpick] Consider using a more descriptive variable or helper function to construct the key for secondary indexes. This can help clarify that the '_index' suffix is intentionally added to differentiate secondary indexes from regular tables or views.

stmt, ok := describedViews[scyllaTable{keyspace: index.KeyspaceName, table: index.Name + "_index"}]

pkg/service/one2onerestore/worker_tgc.go:77

• [nitpick] The parameter 'name' could be renamed to something like 'objectName' or 'tableOrViewName' for clarity, reflecting that this function handles both tables and views.

func (w *worker) getTombstoneGCMode(keyspace, name string, isView bool) (tombstoneGCMode, error) {

[Michal-Leszczynski]

This is not a part of this PR, but still important:

	views, err := w.dropViews(ctx, workload)
	if err != nil {
		return errors.Wrap(err, "drop views")
	}
	defer func() {
		if rErr := w.reCreateViews(ctx, views); rErr != nil {
			err = multierr.Combine(
				err,
				errors.Wrap(rErr, "recreate views"),
			)
		}
	}()

	return w.restoreTables(ctx, workload, target.Keyspace)

Why are we recreating views even if restore tables failed?
Is this some sort of cleanup after failed restore? If so, then it probably should be handled with more care, because if SM restored table partially and the failed, we don't want to wait for the view recreation to finish (such view would need to be re-dropped and re-re-created again anyway). Generally speaking, handling the state of the restored cluster in case of failure is a little bit problematic, so it's okay to just say that the cluster needs to be "nuked" (schema-wise) and restore started from scratch for now.

[Michal-Leszczynski]

nit: it should be more efficient to query all tuples of keyspace_name, view_name, base_table_name at once.