leveltsdb

A TSDB wrapper around eleveldb. It uses eleveldb but provides an interface for timeseries requests instead of just KV.

Usage

Open a DB

1> {ok, Ref} = leveltsdb:open("data").

Write some data, including the DB Ref, the metric name, timestamp, and value

2> leveltsdb:write(Ref, <<"key">>, 1000, 0).
ok
3> leveltsdb:write(Ref, <<"key">>, 1001, 1).
ok
4> leveltsdb:write(Ref, <<"key">>, 1002, 2).
ok
5> leveltsdb:write(Ref, <<"key">>, 1003, 3).
ok
6> leveltsdb:write(Ref, <<"key">>, 1004, 4).
ok
7> leveltsdb:write(Ref, <<"key">>, 1005, 5).

Fold over all <<"key">>s

8> leveltsdb:fold_metric(Ref, <<"key">>, fun({TS, Value}, Acc) -> [{TS, Value}|Acc] end, []).
{1004,4},
{1003,3},
{1002,2},
{1001,1},
{1000,0}]

Fold over a specific TS range for a key

9> leveltsdb:fold_metric(Ref, <<"key">>, 1001, 1003, fun({TS, Value}, Acc) -> [{TS, Value}|Acc] end, []).
[{1003,3},
 {1002,2},
 {1001,1}]

Bucket aggregate values (defaults to 60 seconds). Useful when rolling up values on the fly without having to preaggregate. Supports ranges of timestamps.

10> leveltsdb:aggregate(Ref, <<"key">>, 1000, 1001, <<"avg">>, []).
{ok,[{960,0.5}]}

In this case the mod 60 was 960, the only timestamps for <<"key">> in the range 1000..1001 were 1000 and 1001, with the average value of 0.5.

To set a new bucket size you can pass the option bucket_size

11> leveltsdb:aggregate(Ref, <<"key">>, 1000, 1005, <<"avg">>, [{bucket_size, 2}]).
{ok,[{1000,0.5},{1002,2.5},{1004,4.5}]}

Getting the names of the metrics in the DB

First we store something other than <<"key">>

23> leveltsdb:write(Ref, <<"a">>, 1, 1).
ok
24> leveltsdb:write(Ref, <<"ab">>, 1, 1).
ok
25> leveltsdb:write(Ref, <<"ac">>, 1, 1).
ok
26> leveltsdb:write(Ref, <<"b">>, 1, 1).
ok
27> leveltsdb:write(Ref, <<"c">>, 1, 1).
ok
28> leveltsdb:write(Ref, <<"de">>, 1, 1).

Then we can list all metrics

29> leveltsdb:metrics(Ref).
{ok,[<<"a">>,<<"ab">>,<<"ac">>,<<"b">>,<<"c">>,<<"de">>]}

Or just metrics with a particular prefix

30> leveltsdb:metrics_with_prefix(Ref, <<"a">>).
{ok,[<<"a">>,<<"ab">>,<<"ac">>]}

Each metric name is stored as a separate key in LevelDB so that lookups don't require going over every metric in the DB.