What are the best practices for cluster upgrades in k8s?

[crossoverJie]

Refer to the documentation here:
https://pulsar.apache.org/docs/2.10.x/administration-upgrade/
https://pulsar.apache.org/docs/2.10.x/helm-upgrade/

Here are the steps I took:

1. Upgrade Zookeeper.
2. Disable autorecovery.
3. Upgrade Bookkeeper.
4. Upgrade Broker.
5. Upgrade Proxy.
6. Enable autorecovery.

When a small number of messages are sent during the upgrade process, everything works fine.

But as soon as I increase the number of producers and the rate at which messages are sent, a timeout exception occurs, which results in lost messages.

org.apache.pulsar.client.api.PulsarClientException$TimeoutException: The producer pulsar-test-211-101 can not send 
message to the topic persistent://my-tenant/my-ns/topic-0-partition-8 within given timeout : createdAt 9.171 seconds ago, firstSentAt 2212.751 seconds ago, lastSentAt 2212.751 seconds ago, retryCount 0

Any suggestions to fix this problem?
Thx.