Thursday, June 10th, 06:13 - 08:58 UTC
Android real devices had elevated error rates and some unavailability in our US-West-1 and EU-Central-1 data centers, causing some tests to fail.
Memory spikes were observed in our Android device pools. We also found that two sets of a device pool pod were running, causing some internal endpoint failures.
The remediation here was to reboot the hardware hosting the affected service, which resolved the issue with the multiple pods running. To address the memory spikes, we rolled back to a previously deployed RDC service version.
We are analyzing the code changes that could have caused this to learn how we can avoid this issue in the future.