What rolling updates do #
Rolling updates let ProVide upgrade itself across an entire cluster without taking the service offline. Each Gateway and each node is updated in sequence. Active sessions are allowed to finish gracefully before a component restarts. New connections are routed to whichever components are currently available.
The result: a full upgrade of both the Gateways and the cluster nodes, completed during normal working hours, with no maintenance window and no administrator on standby.
This is the main reason organizations with 24/7 service-level commitments move from a single ProVide Server to a clustered setup with two Gateways.
The two Gateway requirement #
Rolling updates require two ProVide Gateways. With a single Gateway, the service stops while that Gateway restarts to apply its update. With two Gateways, one always remains available to handle traffic while the other is being updated.
The same logic applies to the cluster nodes themselves. ProVide coordinates the sequence so two nodes are never updated at the same time. As long as the cluster has at least two nodes, capacity is preserved throughout.
A cluster with two nodes and two Gateways is the minimum configuration for true zero-downtime updates. More nodes and Gateways add capacity, but the redundancy guarantee is already in place at this minimum.
How a rolling update runs #
When a new ProVide version is published, the update runs in a defined sequence:
1. Gateways are updated first, one at a time. ProVide selects one Gateway. That Gateway is drained: it stops accepting new connections and lets existing sessions complete. The new binary is transferred over the authenticated control channel, the Gateway restarts, and it announces itself as ready. The other Gateway handles all traffic during this window.
2. The second Gateway is updated the same way. Only after the first Gateway is fully back in service does the second Gateway begin its update. At no point are both Gateways unavailable.
3. Cluster nodes are updated next, one at a time. ProVide coordinates the order across the cluster so two nodes are never drained at once. The node about to be updated tells the Gateways to stop sending new connections to it, waits for active sessions to finish, applies the update, and rejoins the cluster.
4. The cycle continues until every node is on the new version. Cluster state, user data, and configuration remain consistent throughout. Synchronization between nodes handles any changes that happened during the update window.
No administrator action is required at any step. The entire sequence runs automatically within the maintenance window you configure.
What clients experience #
From the outside, nothing changes. Clients connect to the same Gateway address before, during, and after the update. New connections are routed to whichever node is currently available. Existing sessions complete on the node they started on.
The control channel between nodes and Gateways uses mutual HMAC authentication, and connections are TLS-protected. The update mechanism itself does not introduce a security exposure: binaries are transferred over the same authenticated channel that handles normal cluster coordination.
Note: Sticky sessions for HTTPS mean a logged-in user stays on the same node throughout a session. If that node is selected for update, the session is allowed to finish before drainage begins. The user is not redirected mid-session.
How to enable it #
Rolling updates are off by default.
To enable:
- Verify your environment meets the requirement: at least two ProVide nodes and two Gateways, all in the same cluster.
- In the ProVide admin interface, open the update settings.
- Enable Rolling Updates and Rolling Gateway Updates.
- Optionally set a maintenance window (default 00:00–04:00) during which updates are permitted to run.
- Save.
Future ProVide updates will then be applied automatically, in the correct sequence, without service interruption.
For the broader picture of clustering and when each level applies, see [What is high availability clustering in ProVide Server?]
For the full setup walkthrough, see [How do I set up an SFTP cluster with load balancing and failover?]
Released in ProVide 2026-04 (Iron). See the release notes for the full change list.
