ragged chart with Max1hDownsampling

[ivmaks]

Hello community! I can not find an explanation for this behavior:

when Max1hDownsampling I observe emptiness

with Max5mDownsampling it seems that everything is fine

curl -s 'https://tq-ww3d-int.wbz.sqtools.ru/api/v1/query_range?query=up%7Binstance%3D%22app56.ww3d-int.wbz.sqtools.ru%3A9102%22%7D&dedup=true&partial_response=true&start=1550479236.65&end=1555922436.65&step=21772&max_source_resolution=1h&_=1555922161796' > 1h.txt
curl -s 'https://tq-ww3d-int.wbz.sqtools.ru/api/v1/query_range?query=up%7Binstance%3D%22app56.ww3d-int.wbz.sqtools.ru%3A9102%22%7D&dedup=true&partial_response=true&start=1550479236.65&end=1555922436.65&step=21772&max_source_resolution=5m&_=1555922161796' > 5m.txt

5m.txt
1h.txt

thanos --version
thanos, version 0.3.2 (branch: HEAD, revision: 4b7320c0e45e3f48a437bd19294f569785bafb02)
  build user:       circleci@4ea007a18ffa
  build date:       20190304-17:16:56
  go version:       go1.10.8

thanos bucket inspect > inspect.log
inspect.log

[bwplotka]

Yea, looks like you don't have 1h downsampling blocks yet and behaviour for fallback to 5m does not really work? (:

[ivmaks]

The problem is reproduced in several environments.

If we remove the 1h samples, the graphics are rendered perfectly. Until the compactor starts and it does not generate 1h samples.

[bwplotka]

I think we are fixing this in #1146

[bwplotka]

Should be fixed by #1146