Mitigating command failures during redis maintenance

[venkatpotru]

We are using lettuce with following configurations,

Lettuce version: 6.2.1, cluster client
redis cluster configuration: 2 masters and 2 slaves, OSS cluster API enabled

topology refresh and other configurations,
autoReconnect: true,
enablePeriodicRefresh: true,
enableAdaptiveRefreshTrigger: ALL,
dynamicRefreshSources: true,
refreshPeriod: 5 seconds,
readFrom: MASTER_PREFERRED,
readFrom: masterPreferred,

Maintenance is done on nodes by following procedure like in https://docs.redis.com/latest/rs/clusters/maintenance-mode/ . During that period, we are seeing different kind of exceptions for 10 to 25 seconds,

1. RedisCommandTimeoutException: command timed out after 5 seconds
2. io.lettuce.core.cluster.PartitionSelectorException: Cannot determine a partition to ...
3. RedisConnectionException: unable to connect ...
4. IOException: Connection reset by peer

Majorly we get 1st and 2nd exception. Looking for advise on mitigating the failures, inputs on following would be great help..

• apart from 1st exception, commands failing with other exceptions is expected behavior? I think commands are not retried by lettuce for those exceptions? Apart from retry from application, is there any other solution to address them?
• read operations are also failing with similar exceptions even though readFrom is 'MASTER_PREFERRED'. Its also expected behavior or are we missing some configuration?