We have two AWS regions: usw2 and use1. We also have two colos with signing servers: mdc1 and mdc2. mdc1 is on the west coast; mdc2 is on the east coast.
Currently, we randomize our signing server list and go down the randomized list to determine which signing servers to use, so we have an equal chance of hitting either colo from either region. Ideally, we should prefer the colo nearer to each region, and fail over to the other: use1 should prefer mdc2, and usw2 should prefer mdc1. We could do this by either providing a primary pool and a backup pool, or by adding priority weighting to each server. (Then we'd need a puppet patch to determine which region the signing scriptworker is in, and apply the appropriate passwords file.)
Related to #68, but neither is a hard blocker for the other.
We have two AWS regions: usw2 and use1. We also have two colos with signing servers: mdc1 and mdc2. mdc1 is on the west coast; mdc2 is on the east coast.
Currently, we randomize our signing server list and go down the randomized list to determine which signing servers to use, so we have an equal chance of hitting either colo from either region. Ideally, we should prefer the colo nearer to each region, and fail over to the other: use1 should prefer mdc2, and usw2 should prefer mdc1. We could do this by either providing a primary pool and a backup pool, or by adding priority weighting to each server. (Then we'd need a puppet patch to determine which region the signing scriptworker is in, and apply the appropriate passwords file.)
Related to #68, but neither is a hard blocker for the other.