Support Forums : DataStax

DataStax Support Forums » Apache Cassandra

Problem with get()

(10 posts) (2 voices)

Started 4 months ago by rajesh.kulkarni2000@gmail.com
Latest reply from rajesh.kulkarni2000@gmail.com
This topic is resolved

Tags:

No tags yet.

rajesh.kulkarni2000@gmail.com
Member

Hi,

I have a ColumnFamily in cassandra with only 4 rows (keys). Each row has a lot of columns (and corresponding values). The 'column names' are numbers but stored as 'UTF-8' strings (for some reasons). There are literally millions of columns per row. I want to do a selective 'get' from each row to fetch values of very small number of columns...something like this (using pycassa python)

from pycassa.pool import ConnectionPool
from pycassa.columnfamily import ColumnFamily
pool = ConnectionPool('BIGDATA_1', server_list=['127.0.0.1'])
col_fam = ColumnFamily(pool, 'row1')
print col_fam.get('IDTOEVENT',['12', '135', '1234'])

The output I get is really strange:-
OrderedDict([(u'12', 'Value of 12'), (u'1234', 'Value of 1234')])
If you see value for column '135' is missing.
If I do a get() on individual column names like '12', '135' and '1234' I see correct results.
I feel, since column names are UTF-8 encoded, they are sorted alphabetically inside cassandra. Hence, while giving the output, Cassandra is throwing 12, followed by 1234 and then perhaps runs 'out of range' or 'out of buffer' limit. This missed 135 ..
I am still new to Cassandra, so there is a high chance that I have missed something.
Could someone please explain what is happening under the hood?

Thanks,
Rajesh Kulkarni

Posted 4 months ago #
thobbs
Administrator

What version of Cassandra are you using? What query are you using to fetch '135' individually, just the same thing without the other two column names?

It looks like you're doing everything correctly, so it's possible that there's a bug here.

Posted 4 months ago #
rajesh.kulkarni2000@gmail.com
Member

Hi thobbs,
Many thanks for your quick response

I am using Cassandra version 1.2.1 (Downloaded apache-cassandra-1.2.1-bin.tar from http://cassandra.apache.org/download/)

And Yes, I am using exactly the same query/get request without other column names. Python code for the same looks like this

print col_fam.get('IDTOEVENT',['12'])
print col_fam.get('IDTOEVENT',['135']) --- Works
print col_fam.get('IDTOEVENT',['1234'])

This gives me the correct output but the get() with multiple column names does not give the result for column '135'
print col_fam.get('IDTOEVENT',['12', '135', '1234'])-- Fails

Any clue why?

Thanks,
Rajesh Kulkarni

Posted 4 months ago #
thobbs
Administrator

Rajesh,

I can reproduce the issue now, thanks for the details. I'll look into the cause a bit more and keep you updated.

Posted 4 months ago #
thobbs
Administrator

I've opened https://issues.apache.org/jira/browse/CASSANDRA-5225 to address the bug.

Posted 4 months ago #
rajesh.kulkarni2000@gmail.com
Member

Thanks thobbs for the investigation! This is a critical feature for me as I have to get() random columns from the the row and hence I would request you to please keep me posted on this thread.

Best Regards,
Rajesh Kulkarni

Posted 4 months ago #
thobbs
Administrator

Rajesh, you can "watch" JIRA tickets if you set up an account, which basically means you will get emailed if anything changes on the ticket. That's probably the easiest way for you to keep track of it.

Posted 4 months ago #
rajesh.kulkarni2000@gmail.com
Member

Hi thobbs,
"Sylvain Lebresne" has provided a fix. We integrated it at our end, built cassandra and it worked pretty well in our test environment.

Since I am new to cassandra community, I wish to ask few more questions, especially relating to the process to be followed
1)How to report a bug to cassandra community? I used this forum and it worked pretty well. Is this the right way Or should I raise a bug/defect in jira directly?
2)Now that the reported defect is fix (thanks to Sylvain Lebresne), what is the next step? Who will commit the code? When will the committed code be available?

I would appreciate if you could throw some light on this.

Thanks,
Rajesh Kulkarni

Posted 4 months ago #
thobbs
Administrator

1)How to report a bug to cassandra community? I used this forum and it worked pretty well. Is this the right way Or should I raise a bug/defect in jira directly?

If you're sure the issue is a bug, you can certainly open a bug in JIRA directly. If you're not sure, feel free to ask a question about what behavior you're seeing here in the forums or on the Cassandra users' mailing list (you can find the info on it at the bottom of this page: http://cassandra.apache.org/).

2)Now that the reported defect is fix (thanks to Sylvain Lebresne), what is the next step? Who will commit the code? When will the committed code be available?

It will be committed by Sylvain or a reviewer of the ticket once it has been fully reviewed and tested. It will be committed to the cassandra-1.2 branch. The fix will be included in the 1.2.2 release, which will probably be released within the next few weeks. Bugfix releases happen about once a month.

I hope that helps! Let me know if you have any other questions.

Posted 4 months ago #
rajesh.kulkarni2000@gmail.com
Member

Thanks thobbs. It certainly helped to clarify my doubts about the process to be followed. I will ask more questions when needed.

Regards,
Rajesh Kulkarni

Posted 4 months ago #

RSS feed for this topic