Confusing _internal.shard numbers - mismatch from expected

Store
1

We have a ton of databases of various sizes in our Influx install, and am trying to get some relative size numbers between them.

Query:

SELECT FIRST("diskBytes") as "diskBytes", FIRST("seriesCreate") as "seriesCreate"
FROM "shard"
WHERE "database" = 'org_1'
GROUP BY "database", "retentionPolicy"
ORDER BY DESC
LIMIT 1

Result:

[{"name"=>"shard",
  "tags"=>{"database"=>"org_1", "retentionPolicy"=>"dbmetrics"},
  "values"=>
   [{"time"=>"1970-01-01T00:00:00Z",
     "diskBytes"=>81626492,
     "seriesCreate"=>21935}]},
 {"name"=>"shard",
  "tags"=>{"database"=>"org_1", "retentionPolicy"=>"autogen"},
  "values"=>
   [{"time"=>"1970-01-01T00:00:00Z",
     "diskBytes"=>72464648,
     "seriesCreate"=>21935}]}]

Mismatch of result!

The size of bytes adds up to: 81626492 + 72464648 = 154091140, or about 154 megabytes.

That result doesn’t match up to what appears on disk. The filesystem reports a drastically different size.

668M    /influx_data/influxdb/influxdb/data/org_1

I haven’t inspected series yet, but 21k is likely significantly low as well.

What is this really measuring?

Can somebody clear up what the shards measurement really is, and if it’s measuring what I think it is? Ideally I want to know the relative expense of the different separate influx databases being hosted on this instance.

Versions:

Influx: version 1.4.0~n201709040800
Ruby client