Yup. There was a GCP outage a couple of years ago like this.
I don’t remember the exact details, but it was something along the lines of a config change went out that caused systems to incorrectly assume there were huge bandwidth constraints. Load shedding kicked in to drop lower priority traffic which ironically included monitoring data rendering GCP responders blind and causing StackDriver to go blank for customers.
I don’t remember the exact details, but it was something along the lines of a config change went out that caused systems to incorrectly assume there were huge bandwidth constraints. Load shedding kicked in to drop lower priority traffic which ironically included monitoring data rendering GCP responders blind and causing StackDriver to go blank for customers.