StatusOutput does not support set(long) when writing a string using the same connection that writes set of strings in transaction

[gigi888]

Hello,

Our AWS lambda uses a single connection to connect to redis 6 with lettuce 6.1.5. It writes a set of strings in transaction.

            RedisCommands<String, String> redisCommands = redisConnection.sync();
            // cache the whole set in one transaction
            redisCommands.multi();
            val.stream().forEach(
                    (String s ) -> { redisCommands.sadd(key, s); }
            );
            redisCommands.expire(key, ttl);
            TransactionResult transactionResult = redisCommands.exec();
            if( transactionResult.wasDiscarded() ) {
                logger.error(" Redis not able to write set of strings for key = " + key);
            }

In the meanwhile, another thread using the same connection, writes a simple string value to another key,

            RedisCommands<String, String> redisCommands = redisConnection.sync();
            redisCommands.set(key2, val);
            redisCommands.expire(key2, ttl);

while load testing this code, I got quite some
java.lang.UnsupportedOperationException: io.lettuce.core.output.StatusOutput does not support set(long)

here and there. I don't have such problem when only writing set. What is the suggested usage of our use case, when we need to write a set of strings with main thread, and just simple string with another thread?

Please help. Appreciate it

[mp911de]

When using transactions, then you must use a dedicated connection, see https://lettuce.io/core/release/reference/#_transactionsmulti.

[gigi888]

Thank you for the the answer!

[hisnamewasjaan]

If using synchronisation on the connection when it is being used in transactions, would this kind of error still be expected?

We are experiencing the same error using lettuce 6.2.6.RELEASE.
We are using a single connection with all transactional code synchronised on the connection. And non-transactional use not synchronised.

Maybe it's just me, but this statement

You should not use transactional commands (MULTI) on shared connection.

from Pipelining-and-command-flushing is maybe a little ambiguous? What exactly does it mean?
Am I mistaken thinking that synchronising on the connection when using MULTI would be sufficient for it be considered not-shared?

Thanks for any clarification.