From 0bbf2808ed02d5a2e44f3a7bb4db6190bdcac87c Mon Sep 17 00:00:00 2001 From: Thomas Jackson Date: Tue, 3 Jul 2018 13:29:27 -0700 Subject: [PATCH] TODO is obsolete. --- TODO | 47 ----------------------------------------------- 1 file changed, 47 deletions(-) delete mode 100644 TODO diff --git a/TODO b/TODO deleted file mode 100644 index 5fc2826c4..000000000 --- a/TODO +++ /dev/null @@ -1,47 +0,0 @@ -Workflow: - - Get query - - Break query up into sub-queries to send - - fetch data for sub-queries - - combine data into result - - return result - -Notes To sort out: - - handle missing data due to broken instance in cluster - - remove need for always aggregating data (even if it isn't ever looked at) - - dynamic discovery of prometheus hosts (adding/removing etc. is "safe"!) - - metrics on query performance and cluster health (as we're querying things!) - - - - -Do we need? - - affinity? - -- if we don't then graphs might move around as the times within the cluster - won't exactly match up-- if we do have affinity config is complicated - as to *breaking* the affinity - - - -Use-cases to cover: - - Mutiple redundant prometheus clusters - -- pick which timeseries to serve (most data, best spread data, etc.) - -- merge datasets together (all unique data points?) - - separate shards of prometheus - -- merge 2 data from N clusters into a single result - -- this requires parsing the query to see if the top level query expr - is an aggregation, if so, then we need to aggregate the vectors - across the clusters that we saw - -Phases: - - Read-Only - -- static host list - -- single query to each cluster - -- merge data within cluster - -- dynamic host list - - Read-Write - -- Presumably a delete would be successful if all hosts in the target return GOOD - - -TODO: - - finish arg implementation of HTTP API - - add "native" step to prometheus API - - make source fetching pluggable (influx, etc.) - -