Consistent Timeouts After "1062" Duplicate Entry Errors

[marwan-at-work]

I have noticed that whenever I get a uniqueness constraints error, the following INSERT/DELETE queries always get timed out.

To reproduce:

1. Create a users table with a unique column called "email"
2. Create a Go db using @fatih 's answer: How do I connect to PlanetScale using Go #24 (comment)
3. Run the following Go code:

func main() {
	// this works
	tx, err := db.Begin()
	must(err)
	_, err = tx.Exec(`INSERT INTO users (email) values ('one')`)
	must(err)
	must(tx.Commit())
	
	
	// force a duplicate entry error
	tx, err = db.Begin()
	must(err)
	_, err = tx.Exec(`INSERT INTO users (email) values ('one')`)
	if err == nil {
		must(fmt.Errorf("expected a duplicate entry error"))
	} else {
		// rollback the transaction
		fmt.Println(tx.Rollback())
	}
	
	// This is where the bug is:
	tx, err = db.Begin()
	must(err)
	_, err = tx.Exec(`DELETE FROM users`)
	must(err) // the error happens here after a 20 second timeout
	must(tx.Commit())
}

func must(err error) {
	if err != nil {
		panic(err)
	}
}

I end up with the following error:

Error 1317: target: gofigdev.-.master: vttablet: (errno 2013) due to context deadline exceeded, elapsed time: 20.000052422s, killing query ID 26989 (CallerID: unsecure_grpc_client) [recovered]

[aquarapid]

Verified manually (using begin/commit/rollback); I did not run into the timeout; but I do see a 15 second or so pause after executing the DELETE from users; you probably hit the timeout because you do this programmatically.

Our engineering team is looking into it.

[mattrobenolt]

This should be fixed as well. Thanks for the report!

[marwan-at-work]

Can confirm 🙂