AWS fsx documentation change
Summary
Added documentation about required downtime during Lustre upgrades and clarified failure behavior when clients remain connected. Added automatic rollback details for failed upgrades.
Security assessment
Changes focus on operational reliability and upgrade procedures, not security vulnerabilities. Mentions of client disconnections are for upgrade success, not security hardening.
Diff
diff --git a/fsx/latest/LustreGuide/managing-lustre-version.md b/fsx/latest/LustreGuide/managing-lustre-version.md index 054886a5b..6892b8398 100644 --- a//fsx/latest/LustreGuide/managing-lustre-version.md +++ b//fsx/latest/LustreGuide/managing-lustre-version.md @@ -39,0 +40,2 @@ We recommend following these best practices before upgrading the Lustre version + * **Plan for file system downtime:** For the upgrade to succeed all clients must be disconnected before the upgrade starts and remain disconnected until the upgrade finishes. Total downtime depends on the time to identify and unmount all clients, the upgrade process itself (typically under 30 minutes for the vast majority of file systems), and the time to remount clients and resume workloads once the upgrade completes. Clients unmount and remount duration varies by your environment and number of clients. + @@ -47 +49 @@ To upgrade your FSx for Lustre file system to a newer version, follow the listed - 1. **Unmount all clients:** Before initiating the upgrade, you must unmount the file system from all client instances accessing your file system. You can verify that all clients are successfully unmounted by using the `ClientConnections` metric on Amazon CloudWatch - this metric should display zero connections. The upgrade process will not proceed if any clients remain connected to the file system. + 1. **Unmount all clients:** Before initiating the upgrade, you must unmount the file system from all client instances accessing your file system. You can verify that all clients are successfully unmounted by using the `ClientConnections` metric on Amazon CloudWatch - this metric should display zero connections. The upgrade process will not succeed if any clients remain connected to the file system. @@ -80,0 +83,2 @@ The following example updates the file system's Lustre version from 2.12 to 2.15 +If the upgrade workflow fails (e.g., if a client is still connected to the file system), the file system is automatically rolled back to its original Lustre version and state. In such a case, the administrative action event will contain a failure message with guidance on how to address the issue before retrying the upgrade. +