Wednesday 23rd January 2019

Service disruption — High I/O wait time

An issue in our distributed File System is currently disrupting I/O-intensive services (slowdown and/or occasional timeout), including the LibreOffice website, AskBot, and the wiki. Update 14:30 UTC: A scheduling race on a node was disrupting data propagation from other nodes, preventing proper syncing and forcing regular healing operations (which in turns drained I/O thus starved the guests). The race was resolved and the I/O wait time is now back to normal.