Hi All, I have a 3 node cluster hosting a Riak TS database. During a query
I run into a 1022 error. The characteristics of my data and other setup is
below. I am trying to understand
1) the error and remedy
2) what are good conf setting for my case.
1 minute interval with 30 columns per minute. Indexed on time (quantum of
1d) and two other columns.
avg(<some column>) where time > t1 and time < t2 and key1=<some val>
and key2 =<someval>
Strangely enough, the query works when t2-t1 <= 5 days; however, for any
duration longer than that I get a 1022 error (projected result of query is
too big). I am using RiakTS v1.5 and it seems in this version this setting
is controlled by riak_kv.query.timeseries.max_quanta_span (set to 5K/(1K
actual?) ); so, is the result purely a coincidence?
It is hard to determine whether this is a coincidence without having the actual dataset. The bug is fixed in
w.r.t 1022 error: Docs say that max_returned_data_size is the parameter I
can tune to fix this issue.
However, it seems this parameter controls the size of the result; so, Is
there any reason I get this error code on a avg calculation?
I have not changed the max_returned_data_size setting to see if it will fix
the issue; just trying to understand whether this is the size of the data
scanned or size of the data returned
Aggregate functions like AVG use the max_returned_data_size setting like a non-aggregate query, though the result is one row. Setting it high enough to hold the query results should allow the query to complete.
Finally, I have 1m interval data for a year and in a production setting I
would like to go to 10s data for 3 years; Are there any recommended value
for quanta I should be using?
I would really appreciate any insight/input into the issue and any
recommendation on how to configure the table.
There is some advice in the docs on quantum size here.