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 Grant Ingersoll <gs...@apache.org> on 2008/11/19 03:13:50 UTC
Wait Flush, Wait Searcher and commit Scenarios
Was wondering if anyone can fill me in on the when and why I would set
waitFlush and waitSearcher to false when sending a commit command? I
think I understand what they do technically (I've looked at the code),
but I am not clear about why I would want to do it. Is there a risk
in setting them to false? Wouldn't my response return a lot faster if
it was set to false? Yet, the default is true for both, so I assume
there is a reason I want to wait...
Thanks,
Grant
Re: Wait Flush, Wait Searcher and commit Scenarios
Posted by rulinma <ru...@gmail.com>.
good.
--
View this message in context: http://lucene.472066.n3.nabble.com/Wait-Flush-Wait-Searcher-and-commit-Scenarios-tp491509p4131651.html
Sent from the Solr - User mailing list archive at Nabble.com.
Re: Wait Flush, Wait Searcher and commit Scenarios
Posted by Yonik Seeley <yo...@apache.org>.
On Tue, Nov 18, 2008 at 10:55 PM, Mark Miller <ma...@gmail.com> wrote:
> Does waitFlush do anything now? I only see it being set if eclipse is not
> missing a reference...
Not currently. The idea was that if waitFlush== false that the call
would be totally asynchronous and return immediately. If
waitFlush==true, then the call would return only after everything was
flushed to stable storage (which is always the case now).
-Yonik
p.s. late replies since I'm getting back from a week of travel.
Re: Wait Flush, Wait Searcher and commit Scenarios
Posted by Mark Miller <ma...@gmail.com>.
Does waitFlush do anything now? I only see it being set if eclipse is
not missing a reference...
Ryan McKinley wrote:
> waitFlush I'm not sure...
>
> waitSearcher=true it will wait until a new searcher is opened after
> your commit, that way the client is guaranteed to have the results
> that were just sent in the index. if waitSearcher=true, a query could
> hit a searcher that does not have the new documents in the index.
>
>
> On Nov 18, 2008, at 9:13 PM, Grant Ingersoll wrote:
>
>> Was wondering if anyone can fill me in on the when and why I would
>> set waitFlush and waitSearcher to false when sending a commit
>> command? I think I understand what they do technically (I've looked
>> at the code), but I am not clear about why I would want to do it. Is
>> there a risk in setting them to false? Wouldn't my response return a
>> lot faster if it was set to false? Yet, the default is true for
>> both, so I assume there is a reason I want to wait...
>>
>> Thanks,
>> Grant
>
Re: Wait Flush, Wait Searcher and commit Scenarios
Posted by Ryan McKinley <ry...@gmail.com>.
I am using waitSearcher=false with a crawler. The crawling thread
finishes a set of stuff, and calls <commit/>. It does not want to
search, it gets back to crawling ASAP
On Nov 18, 2008, at 11:35 PM, Grant Ingersoll wrote:
> That explains true, but what about false? Why would I ever set it
> to false? I f I don't wait, how will I ever know when the new
> searcher is ready?
>
> On Nov 18, 2008, at 10:27 PM, Ryan McKinley wrote:
>
>> waitFlush I'm not sure...
>>
>> waitSearcher=true it will wait until a new searcher is opened
>> after your commit, that way the client is guaranteed to have the
>> results that were just sent in the index. if waitSearcher=true, a
>> query could hit a searcher that does not have the new documents in
>> the index.
>>
>>
>> On Nov 18, 2008, at 9:13 PM, Grant Ingersoll wrote:
>>
>>> Was wondering if anyone can fill me in on the when and why I would
>>> set waitFlush and waitSearcher to false when sending a commit
>>> command? I think I understand what they do technically (I've
>>> looked at the code), but I am not clear about why I would want to
>>> do it. Is there a risk in setting them to false? Wouldn't my
>>> response return a lot faster if it was set to false? Yet, the
>>> default is true for both, so I assume there is a reason I want to
>>> wait...
>>>
>>> Thanks,
>>> Grant
>>
>
>
Re: Wait Flush, Wait Searcher and commit Scenarios
Posted by Grant Ingersoll <gs...@apache.org>.
That explains true, but what about false? Why would I ever set it to
false? I f I don't wait, how will I ever know when the new searcher
is ready?
On Nov 18, 2008, at 10:27 PM, Ryan McKinley wrote:
> waitFlush I'm not sure...
>
> waitSearcher=true it will wait until a new searcher is opened after
> your commit, that way the client is guaranteed to have the results
> that were just sent in the index. if waitSearcher=true, a query
> could hit a searcher that does not have the new documents in the
> index.
>
>
> On Nov 18, 2008, at 9:13 PM, Grant Ingersoll wrote:
>
>> Was wondering if anyone can fill me in on the when and why I would
>> set waitFlush and waitSearcher to false when sending a commit
>> command? I think I understand what they do technically (I've
>> looked at the code), but I am not clear about why I would want to
>> do it. Is there a risk in setting them to false? Wouldn't my
>> response return a lot faster if it was set to false? Yet, the
>> default is true for both, so I assume there is a reason I want to
>> wait...
>>
>> Thanks,
>> Grant
>
Re: Wait Flush, Wait Searcher and commit Scenarios
Posted by Ryan McKinley <ry...@gmail.com>.
waitFlush I'm not sure...
waitSearcher=true it will wait until a new searcher is opened after
your commit, that way the client is guaranteed to have the results
that were just sent in the index. if waitSearcher=true, a query could
hit a searcher that does not have the new documents in the index.
On Nov 18, 2008, at 9:13 PM, Grant Ingersoll wrote:
> Was wondering if anyone can fill me in on the when and why I would
> set waitFlush and waitSearcher to false when sending a commit
> command? I think I understand what they do technically (I've looked
> at the code), but I am not clear about why I would want to do it.
> Is there a risk in setting them to false? Wouldn't my response
> return a lot faster if it was set to false? Yet, the default is
> true for both, so I assume there is a reason I want to wait...
>
> Thanks,
> Grant