You are viewing a plain text version of this content. The canonical link for it is here.

Posted to solr-user@lucene.apache.org by Bernd Fehling <be...@uni-bielefeld.de> on 2015/06/29 17:12:07 UTC

cursorMark and timeAllowed are mutually exclusive?

Hi list,

while just trying cursorMark I got the following search response:

"error": {
    "msg": "Can not search using both cursorMark and timeAllowed",
    "code": 400
}


Yes, I'm using timeAllowed which is set in my requestHandler as
invariant to 60000 (60 seconds) as a limit to "killer searches".

Have nothing found in the ref guides, docs, wiki, examples about this mutually
exclusive parameters.

Is this a bug or a feature and if it is a feature, where is the sense of this?

Regards
Bernd

Re: cursorMark and timeAllowed are mutually exclusive?

Posted by Chris Hostetter <ho...@fucit.org>.

: > Have nothing found in the ref guides, docs, wiki, examples about this mutually
: > exclusive parameters.
: >
: > Is this a bug or a feature and if it is a feature, where is the sense of this?

The problem is that if a timeAllowed exceeded situation pops up, you won't 
get a nextCursorMark to use -- or the one you get might be wrong, and 
could trigger infinit looping.


code doesn't really know about hte cursorMark code, so if a timeAllowed 
"exceeded" siutation pops up, you might not get a nextCursorMark in your 
response, which i considered unacceptible.  if you ask for a cursorMark, 
you get a cursor mark.  if you ask for a cursor mark and include other 
options that make it possible we can't do that, it's an error.

With a bit of work, both could probably be supported in combination -- but 
for now it's untested, and thus unsupported, so we put in that error 
message to make it clear and to guard end users against the risk of 
nonsensical results.

>> Yes, I'm using timeAllowed which is set in my requestHandler as
>> invariant to 60000 (60 seconds) as a limit to "killer searches".

Your best is porbably to confine your cursorMark searches to an alternate 
request handler, not used by your normal arbitrary queries, that doesn't 
have the timeAllowed invariant.



-Hoss
http://www.lucidworks.com/

Re: cursorMark and timeAllowed are mutually exclusive?

Posted by Bernd Fehling <be...@uni-bielefeld.de>.

Hi Chris,

another question pops up, is cursorMark cloud aware?

And if so, who is handling the cursorMark and what if a shard goes down
and comes up again?

Regards
Bernd

Am 30.06.2015 um 08:43 schrieb Bernd Fehling:
> Thanks for your explanation.
> Right out of your head, are there any other options which prevent
> getting a cursorMark?
> 
> Yes, that was also my idea to set up a separate request handler
> for harvesting without timeAllowed.
> 
> As Shawn suggested, a short note about this should go into the documentation.
> 
> Regards,
> Bernd
> 
> 
> Am 29.06.2015 um 18:57 schrieb Chris Hostetter:
>>
>> : > Have nothing found in the ref guides, docs, wiki, examples about this mutually
>> : > exclusive parameters.
>> : >
>> : > Is this a bug or a feature and if it is a feature, where is the sense of this?
>>
>> The problem is that if a timeAllowed exceeded situation pops up, you won't 
>> get a nextCursorMark to use -- or the one you get might be wrong, and 
>> could trigger infinit looping.
>>
>>
>> code doesn't really know about hte cursorMark code, so if a timeAllowed 
>> "exceeded" siutation pops up, you might not get a nextCursorMark in your 
>> response, which i considered unacceptible.  if you ask for a cursorMark, 
>> you get a cursor mark.  if you ask for a cursor mark and include other 
>> options that make it possible we can't do that, it's an error.
>>
>> With a bit of work, both could probably be supported in combination -- but 
>> for now it's untested, and thus unsupported, so we put in that error 
>> message to make it clear and to guard end users against the risk of 
>> nonsensical results.
>>
>>>> Yes, I'm using timeAllowed which is set in my requestHandler as
>>>> invariant to 60000 (60 seconds) as a limit to "killer searches".
>>
>> Your best is porbably to confine your cursorMark searches to an alternate 
>> request handler, not used by your normal arbitrary queries, that doesn't 
>> have the timeAllowed invariant.
>>
>>
>>
>> -Hoss
>> http://www.lucidworks.com/
>>
>

Re: cursorMark and timeAllowed are mutually exclusive?

Posted by Bernd Fehling <be...@uni-bielefeld.de>.

Thanks for your explanation.
Right out of your head, are there any other options which prevent
getting a cursorMark?

Yes, that was also my idea to set up a separate request handler
for harvesting without timeAllowed.

As Shawn suggested, a short note about this should go into the documentation.

Regards,
Bernd


Am 29.06.2015 um 18:57 schrieb Chris Hostetter:
> 
> : > Have nothing found in the ref guides, docs, wiki, examples about this mutually
> : > exclusive parameters.
> : >
> : > Is this a bug or a feature and if it is a feature, where is the sense of this?
> 
> The problem is that if a timeAllowed exceeded situation pops up, you won't 
> get a nextCursorMark to use -- or the one you get might be wrong, and 
> could trigger infinit looping.
> 
> 
> code doesn't really know about hte cursorMark code, so if a timeAllowed 
> "exceeded" siutation pops up, you might not get a nextCursorMark in your 
> response, which i considered unacceptible.  if you ask for a cursorMark, 
> you get a cursor mark.  if you ask for a cursor mark and include other 
> options that make it possible we can't do that, it's an error.
> 
> With a bit of work, both could probably be supported in combination -- but 
> for now it's untested, and thus unsupported, so we put in that error 
> message to make it clear and to guard end users against the risk of 
> nonsensical results.
> 
>>> Yes, I'm using timeAllowed which is set in my requestHandler as
>>> invariant to 60000 (60 seconds) as a limit to "killer searches".
> 
> Your best is porbably to confine your cursorMark searches to an alternate 
> request handler, not used by your normal arbitrary queries, that doesn't 
> have the timeAllowed invariant.
> 
> 
> 
> -Hoss
> http://www.lucidworks.com/
>

Re: cursorMark and timeAllowed are mutually exclusive?

Posted by Shawn Heisey <ap...@elyograg.org>.

On 6/29/2015 9:12 AM, Bernd Fehling wrote:
> while just trying cursorMark I got the following search response:
>
> "error": {
>     "msg": "Can not search using both cursorMark and timeAllowed",
>     "code": 400
> }
>
>
> Yes, I'm using timeAllowed which is set in my requestHandler as
> invariant to 60000 (60 seconds) as a limit to "killer searches".
>
> Have nothing found in the ref guides, docs, wiki, examples about this mutually
> exclusive parameters.
>
> Is this a bug or a feature and if it is a feature, where is the sense of this?

It appears to have been disallowed almost from the beginning of the
cursorMark feature.  It was not present in the first versions of the
patch, but it was already incorporated before anything got committed to SVN.

https://issues.apache.org/jira/browse/SOLR-5463

The reasons for the incompatibility are not clear from the issue notes,
so either hossman or sarowe may need to comment about what makes the two
features fundamentally incompatible, and that info needs to go into the
documentation.

Thanks,
Shawn