For the complete documentation index, see llms.txt. This page is also available as Markdown.

Data Center Replication

The GridGain 8 Data Center Replication dashboard in Enterprise Manager: sender and receiver throughput, replication backlog, and latency per remote data center.

This dashboard shows replication out of and into the cluster: sender throughput and backlog per remote data center, the updates local caches pass to the sender, per-cache queues, and the receiver side. In a bidirectional setup, both the sender and receiver sections show data on each cluster.

Header Tiles

Whether replication is moving, how fast, and how far behind it is. Watch Unacked Entries: it is the backlog of entries sent but not yet acknowledged, and a number that keeps rising means the remote data center has stopped keeping up.

Data Center Replication dashboard header tiles
Panel
Description

Replication State

SENDING or IDLE, based on sender traffic over the last 2 minutes. GridGain doesn't export a replication state metric, so control.sh --dr state remains the authoritative check.

Entries Sent/s

Entries per second sent by this cluster's sender hubs.

Entries Acked/s

Entries per second acknowledged by remote receivers. A persistent gap against Entries Sent/s means a backlog is building.

Entries Failed/s

Entries per second that failed to replicate.

Unacked Entries

Entries sent but not yet acknowledged.

Worst Batch Ack Time

Slowest average batch acknowledgment time across remote data centers.

Pending Queue

Cache updates waiting in the replication pending queue.

Backup Queue

Updates buffered for partitions this node holds as backups.

Sender — Per Remote Data Center

Outbound replication split by destination, so a problem with the link to one data center can be told apart from a cluster-wide problem.

Sender per remote data center section
Panel
Description

Entries Sent / Acked / Failed

Entries sent, acknowledged, and failed per remote data center.

Bytes Sent / Acked / Failed

The same, in bytes.

Batches Sent / Acked / Failed

The same, in batches.

Unacked Entries per Data Center

Entries sent but not yet acknowledged, per destination. Normally near zero; a rising line is the earliest sign of a degraded link.

Average Batch Ack Time per Data Center

Replication latency per destination.

Sender Store & Buffer Backlog

Batches stored for retry and batches held in memory waiting to be sent.

Sender Ingest — From Local Caches

Updates the cluster's data nodes pass to the sender hubs, before batching and sending. Compare with the per-data-center section to tell whether the application is writing less or the link is slow.

Sender ingest from local caches section
Panel
Description

Entries Received from Local Caches

Updates arriving at the sender hubs, per cache.

Entries Filtered & Batches Rejected

Updates the sender discarded, either filtered by the sender group filter or rejected. Both are normally zero.

Replication Backlog — Per Cache

The backlog from the header tiles, broken down per cache. Sustained growth on one cache points to that cache rather than the link.

Replication backlog per cache section
Panel
Description

Pending Queue per Cache

Updates waiting to be sent, per replicated cache.

Backup Queue per Cache

Updates buffered for backup partitions, per cache.

Receiver — Per Remote Data Center

Inbound replication split by source data center, and how long this cluster takes to apply what arrives.

Receiver per remote data center section
Panel
Description

Entries Received / Acked

Entries received and acknowledged per source data center.

Bytes Received / Acked

The same, in bytes.

Batches Received / Acked / In Flight

Batches received and acknowledged, and batches in flight (received but not yet stored and acknowledged).

Receiver Apply Latency & Queue

Time between a batch arriving and being stored, per cache, and the size of the inbound queue in bytes.

This page is: Copyright © 2026 MariaDB. All rights reserved.

spinner

Last updated

Was this helpful?