Phase 6: Cassandra Day-2 Operations After DSE Migration

With the data migrated and DSE decommissioned, the final phase establishes the routine day-2 operations for open-source Apache Cassandra. NodeSync and OpsCenter do not exist for open-source Cassandra, so the repair and management capabilities they provided must be re-established with open-source-compatible tooling before the new cluster is used in production.

This page describes the operational surface to configure. Where AxonOps is referenced, substitute whichever operations platform the team has adopted.

Repair

DSE's NodeSync has no open-source equivalent shipped with Cassandra, so background repair must be configured explicitly. At minimum, schedule a recurring full repair over partitioner ranges across the cluster. The full-repair cadence must complete every token range within gc_grace_seconds (default 10 days) to prevent tombstone resurrection:

nodetool repair -pr --full

A workload-aware scheduler improves on a fixed schedule by adjusting repair intensity to real-time cluster load (throttling during peak traffic and increasing throughput during quiet periods) and by tracking repair coverage per table for visibility into which token ranges have been repaired. AxonOps Adaptive Repair provides this capability as a workload-aware replacement for NodeSync, tracking per-table coverage and alerting on repair failures.

Backup with point-in-time restore

Configure scheduled backups with a retention policy, and enable commitlog archiving for point-in-time restore (recovery to any point, not just the last snapshot). Common storage destinations include the local filesystem, SFTP, Amazon S3, Google Cloud Storage, and Azure Blob Storage. AxonOps Backup provides GUI-driven, scheduled backups with point-in-time restore for open-source Cassandra, supporting the storage destinations above, commitlog archiving for PITR, per-destination retention policies, and tested restores, replacing the OpsCenter backup service after migration.

Test restores before relying on them

A backup that has never been restored is not a verified recovery plan. Test the restore procedure in staging, and verify it end-to-end, before relying on it in production.

Alerting

Configure alert rules for the metrics that matter (read/write latency, pending compactions, GC pause times, disk usage, dropped mutations, and hints) and route them to the relevant incident channels (for example, Slack, PagerDuty, Microsoft Teams, ServiceNow, or email).

When migrating from OpsCenter, many alert rules have direct equivalents; review the curated set and add custom rules for organisation-specific thresholds. Tune thresholds to surface real problems rather than noise. AxonOps monitoring ships curated alert rules for Cassandra out of the box, supports custom rules on any collected metric, and routes alerts through integrations including Slack, PagerDuty, Microsoft Teams, ServiceNow, and email.

Rolling restarts

Automate rolling restarts so that configuration changes and upgrades are applied safely:

• validate cluster health before each node restart;
• wait for each restarted node to rejoin and stabilise before proceeding;
• track progress to identify which nodes have been restarted;
• control restart order, parallelism, and wait conditions.

AxonOps Rolling Restart automates this process, running pre-flight health checks, pacing restarts so each node rejoins and stabilises before the next, and tracking progress.

Service health checks

Schedule application-level checks (for example, custom CQL queries that verify data accessibility and query performance) with defined expected results, and alert when a check fails. These validate the cluster from the application's perspective, beyond infrastructure metrics. AxonOps Service Checks support scheduling these checks with defined expected results and alert on failure, configured through the dashboard without code changes.

Decommission DSE infrastructure

Once the new cluster is fully operational, retire the DSE footprint:

• remove DSE packages from all nodes;
• remove OpsCenter agents and the OpsCenter server;
• clean up DSE-specific configuration files (dse.yaml, DSE security configs);
• update firewall rules;
• update runbooks and on-call documentation to reference the new tooling.

Go / No-Go

Go: all operational features are configured and tested, and the team is trained on the new tooling.

Rollback: not applicable, as DSE has been decommissioned. Ensure backups are in place and tested before reaching this point.

AxonOps for Cassandra covers the full day-2 surface (monitoring, adaptive repair, backup with PITR, alerting, rolling restarts, and service health checks) through a single platform that replaces OpsCenter.

---

Next: Migration is complete. For ongoing operations help, see AxonOps for Cassandra or contact AxonOps.