Re: CQL : Request did not complete within rpc_timeout

2013-02-03 Thread Dave Brosius
If querying by a date inequality is an important access paradigm you probably want a column that represents some time bucket (a month?) And have that column be part of the cql primary key. Thus when a query is requested you can make c* happy by specifying a date bucket to pick the c* row and th

Re: CQL : Request did not complete within rpc_timeout

2013-02-03 Thread Manu Zhang
On Mon 04 Feb 2013 04:42:12 AM CST, Paul van Hoven wrote: Thanks for the answer. Can anybody else answer my other two questions, because my problem is not solved yet? 2013/2/3 Edward Capriolo : This was the issue that prompted the "WITH FILTERING ALLOWED": https://issues.apache.org/jira/browse

Re: CQL : Request did not complete within rpc_timeout

2013-02-03 Thread Paul van Hoven
Thanks for the answer. Can anybody else answer my other two questions, because my problem is not solved yet? 2013/2/3 Edward Capriolo : > This was the issue that prompted the "WITH FILTERING ALLOWED": > > https://issues.apache.org/jira/browse/CASSANDRA-4915 > > Cassandra's storage system can only

Re: CQL : Request did not complete within rpc_timeout

2013-02-03 Thread Edward Capriolo
This was the issue that prompted the "WITH FILTERING ALLOWED": https://issues.apache.org/jira/browse/CASSANDRA-4915 Cassandra's storage system can only optimize certain queries. On Sun, Feb 3, 2013 at 2:07 PM, Paul van Hoven wrote: > I'm not sure if I understood your answer. > >> When you have

Re: CQL : Request did not complete within rpc_timeout

2013-02-03 Thread Paul van Hoven
I'm not sure if I understood your answer. > When you have GB or TB of data any query that adds "WITH FILTERING" > will not work at scale. 1. You mean any query that requires "with filtering" is slow? > Secondary indexes need at least one equality. If you want to do this > at scale you might need

Re: CQL : Request did not complete within rpc_timeout

2013-02-03 Thread Edward Capriolo
Secondary indexes need at least one equality. If you want to do this at scale you might need a different design. Using WITH FILTERING and LIMIT 10 is simply grabbing the first few random rows that match your criteria. When you have GB or TB of data any query that adds "WITH FILTERING" will not wo

Re: CQL : Request did not complete within rpc_timeout

2013-02-03 Thread Paul van Hoven
Okay, here is the schema (actually it is in german, but I translated the column names such that it is easier to read for an international audience): cqlsh:demodb> describe table offerten_log_archiv; CREATE TABLE offerten_log_archiv ( offerte_id int PRIMARY KEY, aktionen int, angezeigt bigin

Re: CQL : Request did not complete within rpc_timeout

2013-02-03 Thread Edward Capriolo
Without seeing your schema it is hard to say, but in some cases "ALLOW FILTERING" might be considered "EXPECT THIS COULD BE SLOW". It could mean the query is not hitting and index and is going to page through large amounts of data. On Sun, Feb 3, 2013 at 9:42 AM, Paul van Hoven wrote: > After fig

CQL : Request did not complete within rpc_timeout

2013-02-03 Thread Paul van Hoven
After figuring out how to use the ">" operator on an secondary index I noticed that in a column family of about 5.5 million datasets I get a rpc_timeout when trying to read data from this table. In the concrete situation I want to request data younger than January 1 2013. The number of rows that sh