You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@cassandra.apache.org by "Sylvain Lebresne (JIRA)" <ji...@apache.org> on 2012/11/06 09:40:14 UTC
[jira] [Created] (CASSANDRA-4918) Remove CQL3 arbitrary select
limit
Sylvain Lebresne created CASSANDRA-4918:
-------------------------------------------
Summary: Remove CQL3 arbitrary select limit
Key: CASSANDRA-4918
URL: https://issues.apache.org/jira/browse/CASSANDRA-4918
Project: Cassandra
Issue Type: Bug
Reporter: Sylvain Lebresne
Assignee: Sylvain Lebresne
Fix For: 1.2.0
Let it be clear however that until CASSANDRA-4415 is resolved, it will put us in a situation where it will be easy to write queries that timeout (and potentially OOM the server). That being said, even with the auto-limit it's not too hard to write queries that timeout if you're not at least a bit careful and so far we've always answer that by saying 'you have to be mindful of how much data your query is asking for'. And while I'm all for adding protection against OOMing the server like suggested by Jonathan on CASSANDRA-4304, I think the arbitrary auto-limit is the worst possible solution to this problem.
Note that until CASSANDRA-4415 is resolved I wouldn't be totally opposed to force people to provide a LIMIT to select queries if we're really thing it will avoids lots of surprise, though tbh I do think it would be enough to just continue to be vocal about the fact that 'you have to be mindful of how much data your query is asking for' and its follow-up 'you should use an explicit LIMIT if in doubt about how much data will be returned'.
But I am *strongly opposed* in keeping the current arbitrary limit because it makes very little sense imo, and the little sense it makes will completely vanish once CASSANDRA-4415 is here, and I don't want to break the API and do a CQL4 to be able to remove that limit later.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-4918) Remove CQL3 arbitrary select
limit
Posted by "Sylvain Lebresne (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-4918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498745#comment-13498745 ]
Sylvain Lebresne commented on CASSANDRA-4918:
---------------------------------------------
I'm clearly fine with that, though while the server-side patch is attached, I'd rather let the cqlsh part to someone that knows the cqlsh code better if possible.
> Remove CQL3 arbitrary select limit
> ----------------------------------
>
> Key: CASSANDRA-4918
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4918
> Project: Cassandra
> Issue Type: Bug
> Reporter: Sylvain Lebresne
> Assignee: Sylvain Lebresne
> Fix For: 1.2.0 rc1
>
> Attachments: 4918.txt
>
>
> Let it be clear however that until CASSANDRA-4415 is resolved, it will put us in a situation where it will be easy to write queries that timeout (and potentially OOM the server). That being said, even with the auto-limit it's not too hard to write queries that timeout if you're not at least a bit careful and so far we've always answer that by saying 'you have to be mindful of how much data your query is asking for'. And while I'm all for adding protection against OOMing the server like suggested by Jonathan on CASSANDRA-4304, I think the arbitrary auto-limit is the worst possible solution to this problem.
> Note that until CASSANDRA-4415 is resolved I wouldn't be totally opposed to force people to provide a LIMIT to select queries if we're really thing it will avoids lots of surprise, though tbh I do think it would be enough to just continue to be vocal about the fact that 'you have to be mindful of how much data your query is asking for' and its follow-up 'you should use an explicit LIMIT if in doubt about how much data will be returned'.
> But I am *strongly opposed* in keeping the current arbitrary limit because it makes very little sense imo, and the little sense it makes will completely vanish once CASSANDRA-4415 is here, and I don't want to break the API and do a CQL4 to be able to remove that limit later.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-4918) Remove CQL3 arbitrary select
limit
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-4918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13500127#comment-13500127 ]
Jonathan Ellis commented on CASSANDRA-4918:
-------------------------------------------
+1 then. Created CASSANDRA-4972 for the cqlsh change. (Would like to get that into 1.2.0 but it's not critical.)
> Remove CQL3 arbitrary select limit
> ----------------------------------
>
> Key: CASSANDRA-4918
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4918
> Project: Cassandra
> Issue Type: Bug
> Reporter: Sylvain Lebresne
> Assignee: Sylvain Lebresne
> Fix For: 1.2.0 rc1
>
> Attachments: 4918.txt
>
>
> Let it be clear however that until CASSANDRA-4415 is resolved, it will put us in a situation where it will be easy to write queries that timeout (and potentially OOM the server). That being said, even with the auto-limit it's not too hard to write queries that timeout if you're not at least a bit careful and so far we've always answer that by saying 'you have to be mindful of how much data your query is asking for'. And while I'm all for adding protection against OOMing the server like suggested by Jonathan on CASSANDRA-4304, I think the arbitrary auto-limit is the worst possible solution to this problem.
> Note that until CASSANDRA-4415 is resolved I wouldn't be totally opposed to force people to provide a LIMIT to select queries if we're really thing it will avoids lots of surprise, though tbh I do think it would be enough to just continue to be vocal about the fact that 'you have to be mindful of how much data your query is asking for' and its follow-up 'you should use an explicit LIMIT if in doubt about how much data will be returned'.
> But I am *strongly opposed* in keeping the current arbitrary limit because it makes very little sense imo, and the little sense it makes will completely vanish once CASSANDRA-4415 is here, and I don't want to break the API and do a CQL4 to be able to remove that limit later.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-4918) Remove CQL3 arbitrary select
limit
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-4918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498171#comment-13498171 ]
Jonathan Ellis commented on CASSANDRA-4918:
-------------------------------------------
How about we remove this server-side, but add it to cqlsh? (Maybe even cut it down to 1K rows there.) That would address my worries about OOMing the server adequately.
> Remove CQL3 arbitrary select limit
> ----------------------------------
>
> Key: CASSANDRA-4918
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4918
> Project: Cassandra
> Issue Type: Bug
> Reporter: Sylvain Lebresne
> Assignee: Sylvain Lebresne
> Fix For: 1.2.0 rc1
>
> Attachments: 4918.txt
>
>
> Let it be clear however that until CASSANDRA-4415 is resolved, it will put us in a situation where it will be easy to write queries that timeout (and potentially OOM the server). That being said, even with the auto-limit it's not too hard to write queries that timeout if you're not at least a bit careful and so far we've always answer that by saying 'you have to be mindful of how much data your query is asking for'. And while I'm all for adding protection against OOMing the server like suggested by Jonathan on CASSANDRA-4304, I think the arbitrary auto-limit is the worst possible solution to this problem.
> Note that until CASSANDRA-4415 is resolved I wouldn't be totally opposed to force people to provide a LIMIT to select queries if we're really thing it will avoids lots of surprise, though tbh I do think it would be enough to just continue to be vocal about the fact that 'you have to be mindful of how much data your query is asking for' and its follow-up 'you should use an explicit LIMIT if in doubt about how much data will be returned'.
> But I am *strongly opposed* in keeping the current arbitrary limit because it makes very little sense imo, and the little sense it makes will completely vanish once CASSANDRA-4415 is here, and I don't want to break the API and do a CQL4 to be able to remove that limit later.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (CASSANDRA-4918) Remove CQL3 arbitrary select
limit
Posted by "Sylvain Lebresne (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-4918?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Sylvain Lebresne updated CASSANDRA-4918:
----------------------------------------
Attachment: 4918.txt
Trivial patch attached
> Remove CQL3 arbitrary select limit
> ----------------------------------
>
> Key: CASSANDRA-4918
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4918
> Project: Cassandra
> Issue Type: Bug
> Reporter: Sylvain Lebresne
> Assignee: Sylvain Lebresne
> Fix For: 1.2.0
>
> Attachments: 4918.txt
>
>
> Let it be clear however that until CASSANDRA-4415 is resolved, it will put us in a situation where it will be easy to write queries that timeout (and potentially OOM the server). That being said, even with the auto-limit it's not too hard to write queries that timeout if you're not at least a bit careful and so far we've always answer that by saying 'you have to be mindful of how much data your query is asking for'. And while I'm all for adding protection against OOMing the server like suggested by Jonathan on CASSANDRA-4304, I think the arbitrary auto-limit is the worst possible solution to this problem.
> Note that until CASSANDRA-4415 is resolved I wouldn't be totally opposed to force people to provide a LIMIT to select queries if we're really thing it will avoids lots of surprise, though tbh I do think it would be enough to just continue to be vocal about the fact that 'you have to be mindful of how much data your query is asking for' and its follow-up 'you should use an explicit LIMIT if in doubt about how much data will be returned'.
> But I am *strongly opposed* in keeping the current arbitrary limit because it makes very little sense imo, and the little sense it makes will completely vanish once CASSANDRA-4415 is here, and I don't want to break the API and do a CQL4 to be able to remove that limit later.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-4918) Remove CQL3 arbitrary select
limit
Posted by "Jonathan Ellis (JIRA)" <ji...@apache.org>.
[ https://issues.apache.org/jira/browse/CASSANDRA-4918?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13498172#comment-13498172 ]
Jonathan Ellis commented on CASSANDRA-4918:
-------------------------------------------
(that is: if querystring does not contain {{limit}}, cqlsh could add on {{LIMIT 1000}} as a default.)
> Remove CQL3 arbitrary select limit
> ----------------------------------
>
> Key: CASSANDRA-4918
> URL: https://issues.apache.org/jira/browse/CASSANDRA-4918
> Project: Cassandra
> Issue Type: Bug
> Reporter: Sylvain Lebresne
> Assignee: Sylvain Lebresne
> Fix For: 1.2.0 rc1
>
> Attachments: 4918.txt
>
>
> Let it be clear however that until CASSANDRA-4415 is resolved, it will put us in a situation where it will be easy to write queries that timeout (and potentially OOM the server). That being said, even with the auto-limit it's not too hard to write queries that timeout if you're not at least a bit careful and so far we've always answer that by saying 'you have to be mindful of how much data your query is asking for'. And while I'm all for adding protection against OOMing the server like suggested by Jonathan on CASSANDRA-4304, I think the arbitrary auto-limit is the worst possible solution to this problem.
> Note that until CASSANDRA-4415 is resolved I wouldn't be totally opposed to force people to provide a LIMIT to select queries if we're really thing it will avoids lots of surprise, though tbh I do think it would be enough to just continue to be vocal about the fact that 'you have to be mindful of how much data your query is asking for' and its follow-up 'you should use an explicit LIMIT if in doubt about how much data will be returned'.
> But I am *strongly opposed* in keeping the current arbitrary limit because it makes very little sense imo, and the little sense it makes will completely vanish once CASSANDRA-4415 is here, and I don't want to break the API and do a CQL4 to be able to remove that limit later.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira