At September 25th, ORES service had higher ~14%) timeout ratio for six hours. Because it ran out space due to too verbose logging.
- Sept 25 10:34:40 UTC 2016: icinga test on ORES failed due to timeout.
- 14:13 UTC: phab:T146581 is created.
- 16:03 The fix deployed in labs.
- 16:26 The fix deployed in prod.
We should have better monitoring disk space and be careful on verbosity of production services logs