feat: Use CLI to perform replace for on-conflict flag

[devanbenz]

No description provided.

[Copilot]

Pull request overview

Moves --on-conflict replace handling into the CLI using a temporary bucket.

Changes:

• Stages restored data before replacing an existing bucket.
• Adds restore API mocks and replacement tests.
• Clarifies CLI flag behavior.

Reviewed changes

Copilot reviewed 4 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
clients/restore/restore.go	Implements staged bucket replacement.
clients/restore/restore_internal_test.go	Tests replacement workflows.
cmd/influx/restore.go	Updates flag help text.
internal/mock/gen.go	Adds restore mock generation.
internal/mock/api_restore.gen.go	Provides generated restore API mock.

Files not reviewed (1)

• internal/mock/api_restore.gen.go: Generated file

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 5 changed files in this pull request and generated no new comments.

Files not reviewed (1)

• internal/mock/api_restore.gen.go: Generated file

Suppressed comments (4)

clients/restore/restore.go:428

• This generated aside name can also exceed InfluxDB's bucket-name limit. Because this rename occurs only after every shard has uploaded, a long but valid original name makes the replace fail at the final swap and leaves the restored data under its temporary name. Use a bounded name that does not include the full original name.

	asideName := fmt.Sprintf("%s-replaced-tmp-%d", swap.name, time.Now().UnixNano())

clients/restore/restore.go:435

• The rollback reuses ctx. The CLI cancels this context on SIGINT/SIGTERM (pkg/signals/context.go:28-30), so if cancellation causes the promotion to fail, the rollback is guaranteed to fail too; the old bucket remains under the aside name and the original name disappears. Use a bounded context detached from cancellation for this compensating rename.

		if _, rollbackErr := c.PatchBucketsID(ctx, swap.oldID).PatchBucketRequest(api.PatchBucketRequest{Name: &swap.name}).Execute(); rollbackErr != nil {
			log.Printf("WARN: Failed to restore original name of bucket %q (currently %q): %v\n", swap.name, asideName, rollbackErr)
		}

clients/restore/restore.go:418

• InfluxDB bucket names must be fewer than 64 characters, but this appends roughly 32 characters to an already-valid name. Any existing bucket with a moderately long name will therefore fail before staging starts. Generate a bounded temporary name independent of the original name.

This issue also appears on line 428 of the same file.

	bkt.BucketName = fmt.Sprintf("%s-restore-tmp-%d", bkt.BucketName, time.Now().UnixNano())

clients/restore/restore.go:371

• If shard upload fails because the CLI context was canceled, this cleanup request reuses the already-canceled context and fails immediately, leaving a potentially full-size temporary bucket behind. Run the compensating delete with a bounded context detached from cancellation while preserving context values.

This issue also appears on line 433 of the same file.

		if err := c.DeleteBucketsID(ctx, bucketID).Execute(); err != nil {
			log.Printf("WARN: Failed to delete temporary bucket %q: %v\n", bkt.BucketName, err)
		}

[Copilot]

Pull request overview

Copilot reviewed 4 out of 5 changed files in this pull request and generated no new comments.

Files not reviewed (1)

• internal/mock/api_restore.gen.go: Generated file

Suppressed comments (2)

clients/restore/restore.go:436

• The same unbounded suffix issue occurs when renaming the old bucket aside: a valid long bucket name makes this PATCH fail, so the fully uploaded temporary bucket can never be promoted. Use the same bounded ID/nonce-based naming helper for the aside name.

	asideName := fmt.Sprintf("%s-replaced-tmp-%d", swap.name, time.Now().UnixNano())

clients/restore/restore.go:356

• Appending the staging suffix to the user’s bucket name can exceed InfluxDB’s bucket-name length limit even though the original name is valid, causing --on-conflict replace to fail before any shards are uploaded. Generate a bounded temporary name independent of the full bucket name (for example, from the existing bucket ID plus a nonce) and cover a maximum-length bucket name.

This issue also appears on line 436 of the same file.

			bkt.BucketName = fmt.Sprintf("%s-restore-tmp-%d", bkt.BucketName, time.Now().UnixNano())