Do you have equality predicated on partition key?

not doing partition on key currently. only simple partition on kafka topic

can you share link to replica group implementation on realtime tables?

Ok could you elaborate why you think your query should not go to all servers?

actually its giving high latency on broker compared to less servers/replicas

And @User clearly we see that server is responding with 1to 5 ms but broker is taking time to respond..

Is there any other solution?

ReplicA group and partitioning will help

is there a sample response and lets move to #C011C9JHN7R

Sure we will post all details there..

Hi, we found an issue where non-pql queries that have grouping columns in the select list which are not included in the group by clause are just ignored instead of throwing a syntax error. Example:

select foo, count(*), max(bar) from table

This query will return results as if it was written:

select count(*), max(bar) from table

i.e. the grouping column is ignored

@User looks like calcite does not support that syntax and requires explicit group by

It doesn't seem to throw an error though, just ignores the group by column. This can be verified in the pinot controller ui.

I am using pinot-3.0, i.e. the last commit where the version is

0.3.0-SNAPSHOT

yes, standard SQL compliant behavior is to have the group by column explicitly in the select list

Otherwise it won't be projected in the result set

note that group by is still done correctly -- it's just that column won't be part of result columns projected

@User lets create an issue, we should be able to support the short hand

@User this is a different case rt, there is no group by clause in the query

@User why are we calling it a grouping column when there is no group by in the query?

presto supports that syntax, i have seen shasta also supports that

its a shorthand 🙂

grouping column - i.e. if the user inadvertently forgot to include a group by. Presto would return an error, more of a question. What's shasta?

yeah, we should return an error for now

yes it should be a syntax error

Yep, just verified using presto's tpch connector:

presto:tiny> select name, count(*) from nation;
Query 20200511_170334_11208_ru3kx failed: line 1:8: 'name' must be an aggregate expression or appear in GROUP BY clause
io.prestosql.spi.PrestoException: line 1:8: 'name' must be an aggregate expression or appear in GROUP BY clause