Short answer: Qumulo Continuous Replication copies the contents of a directory to another cluster. SMB shares, NFS exports and quotas are cluster configuration and are not replicated. If they are missing or outdated on the secondary, clients cannot reach the data after a failover. Keep them in sync continuously, mapped to the replicated paths, and in the same direction as replication.
What replication copies and what it doesn’t
| Object | Replicated with the data? | Needed on the secondary for failover |
|---|---|---|
| Files, directories, file and directory permissions | Yes | – |
| SMB shares incl. share permissions and share options | No | Yes |
| NFS exports incl. access rules | No | Yes |
| Directory quotas | No | Yes, to keep limits after failover |
| Floating IPs and networks | No | Moved during failover |
The gap shows up at the worst moment. Failover works on the storage side, the data is writable on the secondary, and users still can’t connect, because the share created on the primary last month never made it to the DR cluster.
Why this drifts
Shares, exports and quotas change far more often than DR plans are reviewed:
- New projects get new shares, often created by different admins.
- Share permissions change when teams change.
- Quotas are raised when a department runs full.
- Exports get new client networks.
Every change has to be repeated on the secondary, with the target path instead of the source path. One missed change is enough to break access after failover.
Option 1: maintain the secondary by hand
For each change on the primary, the same change on the secondary, in the Web UI or with qq. This works for a handful of shares that rarely change. Make it part of the change process, and compare both clusters regularly, for example by exporting qq smb_list_shares and qq nfs_list_exports from both and diffing the output after mapping the paths.
Option 2: script it
A sync script reads the configuration from the primary and writes it to the secondary. Plan for these details:
- Path mapping.
/Mediaon the source is/qumulo-prod/Mediaon the target. Every share, export and quota has to be translated. - Scope. Sync only objects whose path lies inside a replicated directory. Everything else belongs to the secondary itself.
- Create, change and delete. A share deleted on the primary has to be deleted on the secondary too, otherwise stale shares accumulate.
- Dependent settings. A share is more than its name: share permissions, access-based enumeration, encryption and host restrictions belong to it. An export comes with its access rules.
- Direction during failover. While the secondary is active, admins make changes there. These must flow back to the primary at failback, not be overwritten by the next sync run.
- Cross-replication. If both clusters replicate directories to each other, the script must sync in both directions, per directory.
- Logging. You need to see what the script changed and when, to answer “why is this share different?”.
In the current qq CLI, the relevant commands are qq smb_create_share, qq smb_modify_share, qq nfs_create_export and qq quota_create_quota, or the matching REST endpoints. Older Qumulo Core versions use different names for some of them, so check the command guide for your version.
Points 5 and 6 are what make most home-grown scripts fragile: the sync has to know the current replication direction.
Option 3: automate it with failover awareness
The requirement follows from the list above: configuration should follow the data. Wherever replication currently copies a directory, the shares, exports and quotas for that directory should follow in the same direction.
How yuzuy automates this
yuzuy Synchronization keeps SMB shares, NFS exports and quotas identical between two Qumulo clusters. You choose which of the three object types to sync. Share permissions, NFS access rules and share options belong to their share or export and are always synced with it.
yuzuy detects which directories are part of a replication relationship and only syncs objects inside them, with the paths mapped to the target. Before the first run, a preview shows what yuzuy would create or change, without touching the clusters. After that, changes are synced at a configurable interval, commonly 10 to 60 seconds, and the history shows every change.
Synchronization is failover-aware. Configuration always follows the direction of replication. Changes made on the secondary during failover are synced back to the primary at failback. Clusters that replicate to each other in both directions are supported.
Together with Failover and Failback, this means the secondary is ready whenever you switch, and a failover test doesn’t start with an hour of comparing shares.
Synchronization · How to fail over Qumulo Continuous Replication for SMB and NFS clients
FAQ
Does Qumulo replicate SMB shares and NFS exports? No. Continuous Replication copies directory contents. Shares, exports and quotas must be created on the target cluster separately.
Can a share point to a replication target directory? Yes. The directory is read-only while replication runs, so clients can only read. After failover, the target becomes writable and the share works as normal.
What happens to share changes made on the DR cluster during a failover? Without automation they are lost or overwritten unless someone repeats them on the primary. yuzuy syncs them back at failback.
Sources
- Creating and managing a Continuous Replication relationship (Qumulo)
- qq smb_create_share (Qumulo)
- qq smb_modify_share (Qumulo)
