Recently, we are experiencing this error when trying to access the dashboard in staging:

upstream connect error or disconnect/reset before headers. reset reason: connection termination

According to SRE, this happens only when accessing /p/admin/dashboard and it's most probably due to a connection timeout. The screen tries to load much data and times out before finishing loading, a cache stores the data so next time the screen loads faster. After a period of time, the cache expires and the error happens again.

Attached are some logs from openshift