incubator-cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Pavel Kirienko <pavel.kirienko.l...@gmail.com>
Subject select count(1) and OOM
Date Fri, 05 Jul 2013 17:46:29 GMT
Hi everyone,

I was playing with a single-node Cassandra installation when discovered
that a request like [SELECT COUNT(*) FROM CF] seems to load the entire
dataset of CF into RAM. I am not sure is it expected to behave this way or
not. I'd expect it to iterate through the entire set of rows rather than
collect values in memory.

My steps:

create table big_table (
k int primary key,
idx bigint,
val ascii,
ts timestamp
);
create index on big_table (idx);

I filled the table above with 400 random rows, where column 'val' was
written with random strings of 10MB each. Thus I came up roughly with 4GB
of data.

At this point everything is fine, response delays are pretty good and
memory consumption is adequate.

Things go bad with a counting request like [SELECT COUNT(1) FROM big_table]
- that makes the database die with OOM. However, it is possible to fetch
any column except the huge one: [SELECT k FROM big_table] - this works okay.

As far as I understand, a counting request works roughly the same way as
[SELECT * FROM] with only difference that it doesn't return any data back.
Is my reasoning correct?

Thanks in advance,
Pavel.

Mime
View raw message