cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sylvain Lebresne (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-9802) Better page size for aggregates in cqlsh
Date Tue, 14 Jul 2015 17:05:09 GMT


Sylvain Lebresne commented on CASSANDRA-9802:

bq. If the page size is lower than a fix threshold (e.g. 5000), the page size specified by
the client is ignored and a page size of 5000 is used.

Sure, that's simple, but that's a complete hack. There is no reason to ignore clients page
size below a certain threshold in general. What if the use knows the aggregated values are
huge and a very low page size *is* the right thing in that very specific case?

bq. It is doable but it will add a lot of logic to cqlsh.

Don't know. Only reason I created this is that is that it sounded not all that hard from what
Tyler said on CASSANDRA-9724. And I though cqlsh was already parsing queries for the sake
of completion anyway. But If it's a mess to do, then I'm fine with not bothering, it's not
like using a too low page size in cqlsh is a big deal. That issue is really just a nice-to-have.
And it's not in any way a priority btw.

> Better page size for aggregates in cqlsh
> ----------------------------------------
>                 Key: CASSANDRA-9802
>                 URL:
>             Project: Cassandra
>          Issue Type: Improvement
>            Reporter: Sylvain Lebresne
>            Assignee: Benjamin Lerer
>            Priority: Minor
>             Fix For: 3.x
> As discussed in CASSANDRA-9724, cqlsh uses a small page size (100) which makes sense
for "normal" queries since you don't want to flood the terminal, but is quite sub-optimal
for aggregates where the page size is used as the internal page size for the aggregate (there
is thus no terminal flooding concerns and a page size of 100 is too low).
> We should thus make cqlsh use a bigger page size (stick with the python driver default
for instance) when dealing with aggregate queries. According to Tyler:
> bq. it's not trivial, but it is possible. It should take a day or less of work.

This message was sent by Atlassian JIRA

View raw message