-
Type: Improvement
-
Resolution: Fixed
-
Priority: Major - P3
-
None
-
Affects Version/s: None
-
Component/s: Atlas
-
Labels:None
Description
In Replica Set Tag Descriptions, we describe nodeType as having possible values of ELECTABLE, READ_ONLY, and ANALYTICS. We should describe what each of these types means explicitly. Notably that nodeType:READ_ONLY means a non-electable, read-only node as configured in the Atlas UI.
Why is there confusion? My initial thought when I saw these node types is that an Analytics node is used for analytics and therefore would have the nodeType:ANALYTICS applied to it. This is correct. An Analytics node cannot be written to (because it is not the primary nor can it ever be primary) and therefore should also have nodeType:READ_ONLY applied to it. This is incorrect. By this line of reasoning, I had also expected electable secondaries to also be marked with nodeType:READ_ONLY. Thinking it through, it doesn't make sense for Atlas to reconfigure tag sets based on election outcomes, but without knowing the inner workings of tag sets, it doesn't seem unreasonable at first glance for electable secondaries to be tagged with nodeType:READ_ONLY.
In summary we should explicitly call out the following:
nodeType:ELECTABLE means an "electable node for high availability" as configured in the Atlas UI.
nodeType:READ_ONLY means a "read-only node for optimal local reads" as configured in the Atlas UI.
nodeType:ANALYTICS means an "analytics node for workload isolation" as configured in the Atlas UI.