Schema read failed when one node is shuting down

[jbarotin]

Hi,

We have a cluster of three bare-metal servers (32GB ram, 16 vcore), every server handles Pulsar, Bookkeeper and Zookeeper. We are using Avro encoding of messages.

We ran a test of shutting down a server to check if HA of our cluster is OK.

When the server was down, we received the following exception on client :

Wrapped by: org.apache.pulsar.shade.com.google.common.util.concurrent.UncheckedExecutionException: org.apache.pulsar.shade.org.apache.commons.lang3.SerializationException: Failed at fetching schema info for 1 [pulsar-external-listener-54-1]

We turned on back services, and the exception disappeared.

I checked the source code :
https://github.com/apache/pulsar/blob/e8fedb16ca6b8b02b4981b325899000f1c828395/pulsar-broker/src/main/java/org/apache/pulsar/broker/service/schema/BookkeeperSchemaStorage.java#L595C60-L595C82

And our Pulsar configuration :

managedLedgerDefaultEnsembleSize=2
managedLedgerDefaultWriteQuorum=2
managedLedgerDefaultAckQuorum=2

Any idea why client schema read is not working ? At least, can somebody give me a CLI to check which ledger is storing the Avro schema ?