Knowledge base

Three counters that belong in a total and are not in one verifiedcritical

strange is counted by nothing, unassignable by one total, expired and norecipients by the other — so no number on the screen is the count of failed messages.

In detail

Set out plainly, because it decides what our charts may claim. Thirteen error counters exist. failures takes eleven of them, leaving out unassignable and strange. allerrors takes seven, leaving out everything temporary plus expired, norecipients and strange. No combination of the two gives the total, and strange — the bucket for failures nobody understands — is in neither. The practical size of the error is small on a healthy node and exactly the wrong size on a broken one: an address misconfiguration shows up in unassignable, and an unexplained problem shows up in strange, and both are the cases where the failure chart stays flat. Our rule already covers this — the parts must sum to the whole, and the coverage must be stated — so our failure total is the sum of all thirteen, with the console's two totals available as named subsets for anyone comparing screens.