The past couple of years we have been hoping to have Thumbor running on Kubernetes, bad sadly this hasn't happened yet. We have though been buying replacement servers which we have been adding to our mediawiki clusters (api, app, jobrunner).

Since it is unclear when Thumbor is going to move to kubernetes, so our options are:

• Consider if it makes sense to refresh the servers now (Q4 or Q1 FY21-22),
• keep things as they are until Thumbor gets migrated to kubernetes. Any servers that die along the way, we can replace by stealing machines from the mw* clusters

Notes:

• some thumbor servers have been reporting memory errors
• servers were purchased between 2014-2016
• we are OK in terms of performance, apart from some minor incidents that pop up from time to time