[D] GEO Replication in Pulsar under Network Restrictions [pulsar]

["jichmi (via GitHub)" <gi...@apache.org>]

GitHub user jichmi created a discussion: GEO Replication in Pulsar under Network Restrictions

We are setting up Pulsar clusters across multiple data centers (DCs) with restricted network connectivity. The network can be simplified as three DCs: A, B, and C. Networks are connected between A-B and A-C, but B-C cannot connect directly.
Problem:
After configuring GEO replication assuming full connectivity, we encountered:
1、Frequent replication errors in logs, such as:
`
2025-03-28T09:34:09,655+0000 [pulsar-io-4-6] WARN  org.apache.pulsar.client.impl.ConnectionPool - Failed to open connection to c-dc.com/<unresolved>:6650 : io.netty.channel.ConnectTimeoutException: connection timed out after 10000 ms: c-dc.com/10.xx.xx.xx:6650
2025-03-28T09:34:09,655+0000 [pulsar-io-4-6] WARN  org.apache.pulsar.broker.service.AbstractReplicator - [persistent://global/public/test-topic-partition-0 | b-pulsar-->c-pulsar] Failed to create remote producer (org.apache.pulsar.client.api.PulsarClientException: java.util.concurrent.CompletionException: io.netty.channel.ConnectTimeoutException: connection timed out after 10000 ms: c-dc.com/10.xx.xx.xx:6650), retrying in 57.569 s
`
2、Storage in B and C cannot be released due to unupdated replication cursors until we run:
`bin/pulsar-admin persistent skip-all persistent://global/public/test-topic -s pulsar.repl.c-pulsar`
Question:
Are there better approaches to handle GEO replication in partially connected networks?

GitHub link: https://github.com/apache/pulsar/discussions/24137

----
This is an automatically sent email for commits@pulsar.apache.org.
To unsubscribe, please send an email to: commits-unsubscribe@pulsar.apache.org