summarize function returning incorrect results

[woodsaj]

If i send a request for 3months of data, summarized into 30day averages

target: summarize(metrictank.aggstats.1-mtqa.tank.metrics_active.gauge32, '30day', 'avg', false)
from: 1517414400
until: 1525104059
format: json

i get back

[  
  {  
    "target":"summarize(metrictank.aggstats.1-mtqa.tank.metrics_active.gauge32, \"30day\", \"avg\")",
    "tags":{  
      "name":"metrictank.aggstats.1-mtqa.tank.metrics_active.gauge32",
      "summarize":"30day",
      "summarizeFunction":"avg"
    },
    "datapoints":[  
      [  
        66.50059255122777,
        1516320000
      ]
    ]
  }
]

As well as there being only 1 datapoint returned instead of the expected 3, the timestamp of the returned datapoint is for 12days before the requested From.

If a add a keepLastValue() function to the query, forcing the query to be handed off to Graphite, i get the correct results.

This looks to be a fairly recent bug, as a number of queries that were working last month are not working today.

[Dieterbe]

This looks to be a fairly recent bug, as a number of queries that were working last month are not working today.

native summarize() support in MT was merged on march 8 via #837, likely there is a bug in it

[DanCech]

It seems like the issue may be in the calculation of numPoints, the go code has it as the number of expected output points, where in the python code it's the number of input points.

https://github.com/grafana/metrictank/blob/master/expr/func_summarize.go#L90

https://github.com/graphite-project/graphite-web/blob/master/webapp/graphite/render/functions.py#L4789-L4794

[shanson7]

Also, should it be end-start rather than start-end in the go code?

[shanson7]

Should summarize be resetting the consolidation function?
https://github.com/grafana/metrictank/blob/master/expr/func_summarize.go#L36

Seems like the python version doesn't do this.

[Dieterbe]

Should summarize be resetting the consolidation function?
https://github.com/grafana/metrictank/blob/master/expr/func_summarize.go#L36
Seems like the python version doesn't do this.

yes, see https://github.com/grafana/metrictank/blob/master/expr/NOTES#L73

[Dieterbe]

FWIW interval is 10:

mt-index-cat --prefix metrictank.aggstats.1-mtqa.tank.metrics_active.gauge32 cass -timeout 30s -keyspace \$MT_CASSANDRA_IDX_KEYSPACE -hosts \$MT_CASSANDRA_IDX_HOSTS '{{.Id}} {{.OrgId}} {{.Name}} {{.Interval}} {{.LastUpdate}} {{.Partition}}\n'"
1.2c908409650e626115cf8bf5c74331a0 1 metrictank.aggstats.1-mtqa.tank.metrics_active.gauge32 10 1528101600 4