cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Michael Mior (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-8859) CollationController not using collectTimeOrderedData
Date Tue, 24 Feb 2015 21:58:05 GMT


Michael Mior commented on CASSANDRA-8859:

That is, in Cassandra 2.1.2. I understand that the scenario I mentioned above will use this
code path in 3.0.

> CollationController not using collectTimeOrderedData
> ----------------------------------------------------
>                 Key: CASSANDRA-8859
>                 URL:
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>            Reporter: Michael Mior
>            Priority: Minor
> I'd appreciate some help with a Cassandra 2.1.2 issue I'm experiencing. I'm running a
query which looks like this:
> {code}
> CREATE TABLE single_row_fetch (id uuid PRIMARY KEY, data text)
> SELECT data FROM single_row_fetch WHERE id = ?
> {code}
> When writing test data into this table, I disabled compaction. I then wrote data for
performed a flush, and then overwrote the data, and so on. I varied the number of times the
data was overwritten and flushed. This has the effect of controlling the number of SSTables.
However, given that the table only has one non-key row, only a single SSTable will ever have
the most recent data for this row. I confirmed that the expected number of SSTables were generated
and the timestamps of the  are as expected.
> However, when I run the query with tracing, I see that Cassandra still reads from ALL
of the SSTables via {{collectAllData}} in {{CollationController}}. Given that this query only
fetches a single column, I would expect this query to take the {{collectTimeOrderedData}}
code path and then only examine the first SSTable after seeing that it contains the relevant
> Any insights on why this is the case and it what situations I would get the expected
behaviour would be incredibly helpful!

This message was sent by Atlassian JIRA

View raw message