Allowed downtime for maintenance windows

Joseph Foxton asked on the farmers chat: "We have a question regarding scheduled maintenance windows and token earnings, as we have some exciting opportunities knocking on our door.

In order to provide performant S3 for our customers, we need to upgrade to 10Gb bandwidth in our Salzburg data center, and install a reverse proxy tool. This will require some hours of downtime.

Currently, the connection token earning policy gives us 1 day + 19 hours of downtime per year. Is there a way for us to host a scheduled maintenance outage to perform these upgrades, without eating into our allotted downtime?"

Good and valid questions. Maintenance windows in the centralized datacenter world are common. Most datacenter, co-location, dedicated server providers and even some hosting providers have this and this is where they perform “facility” maintenance affecting a large number of customers. Most service providers, SaaS, FaaS don’t have them as they have built there architecture to not be affected by facility maintenance.

In the TF Farming model a farmer is required to achieve the (external) uptime guarantee (99.5%) in order to have tokens minted and received. There is no concept of a maintenance window in the TF tokonomy. TF Tech has developed ZeroOS in such a way that a new version of the software does not require a downtime, even without affecting existing workloads on nodes. So for the actual operation of the TF Grid there is no necessity to have maintenance windows as the technology (without bugs :wink: does not need them.

For Joe’s specific case I would think the following workaround may work.

  • core connectivity is not going to be interrupted for longer than 3.6 hours (which is the 99.5%) uptime guarantee.
  • the old networking setup can remain operational up to the point where the new one is built and operational.
  • when all is done it is a simple matter of switching a few cables with a minimal amount of downtime, not effecting tokens earnings