Why architecture matters here

Replication fails on lag + broken peer. Architecture matters because WAL + serial + throttle compose.

Advertisement

The architecture: every piece explained

The top strip is source. Source RS. WAL edits. Replication endpoint. Peer cluster.

The middle row is controls. Serial replication. Filters. Bandwidth throttle. Sync vs async.

The lower rows are ops. Failover. Metrics. Ops — peer add + monitor + backfill.

HBase replication v2 — peer + WAL + serial + asynccross-cluster copy of editsSource RSregion serverWAL editsordered logReplication endpointship to peerPeer clustersink RSSerial replicationorder preservedFiltersper-CF or peerBandwidth throttleper peerSync vs asynctradeoffFailoverpeer down handlingMetricslag + queue sizeOps — peer add + monitor + backfillorderfilterthrottlesyncresumewatchwatchoperateoperate
HBase replication v2 across peer clusters.
Advertisement

End-to-end flow

End-to-end: RS writes edit to WAL. Replication reads WAL, ships to peer. Peer applies. Serial mode ensures order per region. Lag monitored; peer down triggers queue growth alert.