SMB fail-over scenarios and upgrade

When you are planning an IBM Storage Scale system configuration that has a high number of SMB connections, you must consider the impact that a fail-over can have on the performance of the system.

When an IBM Storage Scale protocol node fails, the IP addresses that are hosted by that protocol node are moved to another IBM Storage Scale protocol node. SMB clients must reconnect to one of the remaining CES nodes. After the fail-over is processed, the same IP address can be used. The remaining protocol nodes handle all the SMB connections.

Therefore, when you plan an IBM Storage Scale system that has a high number of SMB connections, some buffer, in terms of number of SMB connections, must be factored into the overall system configuration. This contingency buffer prevents a system overload during a fail-over scenario, thus reducing any adverse effects on the system performance.

A similar consideration applies to SMB upgrades. The SMB upgrade happens in two phases. During the first phase of the upgrade process, the first half of the nodes are updated. The remaining nodes handle the SMB connections of the updated node. When the first half of the nodes are updated, the upgrade moves to the second phase. In the second phase, SMB is shut down completely. This practice is followed to update the SMB code on all of the remaining protocol nodes concurrently. However, it leads to a brief outage of the SMB service. For more information, see Upgrading SMB packages.