Question 1082

Paging on the Cause

One degraded cache node causes 43 pages in ten minutes: high CPU on six hosts, replication lag, queue depth, thread pool saturation, and more. On-call has started muting channels. What is the right restructuring of these alerts?

A

Route all of the alerts into one channel and have on-call triage the batch by hand, escalating only the ones that appear to be user-visible in impact

B

Page on a small number of user-visible symptoms tied to SLOs, and demote the cause-level signals to dashboards or tickets used during diagnosis

C

Keep all 43 of the alert rules but raise each threshold by roughly 20 percent, so only conditions well past normal variation fire a page

D

Add a dependency-aware alert per component so that responders can read the root cause straight off the page instead of diagnosing it live