Shard Reboots

[habiff]

Enterprise datacenters have regular maintenance windows, with timing guarantees. Therefore I have no doubt that a counter could be created, timing the window for ryzom.

During the window, no guarantees. Outside of the window, standard SLA applies.

Expectations and bad reviews could be mitigated by announcing the maintenance windows, and ensuring the length of the window is twice as long as the installation, and rollback of any applied changes (essentially four times the needed time). This means testing in dev or release candidate environments prior to production launch, to understand the necessary time required to implement and back out any change.