Skip to content

2025-08-26 19:25:00Z WAL Saturation Event

image

source

This one is most puzzling because I have observed the following for the time frame of the incident:

  1. There is no abnormal write activity on any tables.
  2. There are not an abnormally large number of autovacuum workers running.
  3. There was a burst of disk write activity on the primary.

My working theory is that we had a low number of deletes on a table that has huge tuples, which would explain the write burst and sustained WAL volume while also flying under the radar while trying to isolate the busiest table for write traffic.