We're going to take one more measure to help performance, which may cause a short hitch.

Storage has been restarted. You might notice suboptimal performance still until the cache is warmed up. We're monitoring that.

Last night, we transferred storage from CephFS to NFS for the 'http.pca' cluster, as announced on January 9. We are seeing a performance drop since, which we can only resolve by restarting storage. This will cause sites to be unavailable for 10 minutes.

We are sorry for the short-term announcement. We weighed the risk of 10 minutes of downtime during the day against suboptimal performance until tonight.