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 John Blythe <jo...@curvolabs.com> on 2017/08/08 13:02:43 UTC

missing documents after restart

hi all.

i have a core that contains about 22 million documents. when the solr
server is restarted it drops to 200-400k. the dashbaord says that it's both
optimized and current.

is there config issues i need to address in solr or the server? not really
sure where to begin in hunting this down.

thanks-

Re: missing documents after restart

Posted by John Blythe <jo...@curvolabs.com>.
haha, can't have that now!

-- 
*John Blythe*
Product Manager & Lead Developer

251.605.3071 | john@curvolabs.com
www.curvolabs.com

58 Adams Ave
Evansville, IN 47713

On Fri, Aug 11, 2017 at 2:44 PM, Erick Erickson <er...@gmail.com>
wrote:

> Thanks for closing this out, I was breaking out in hives ;)
>
> Erick
>
> On Fri, Aug 11, 2017 at 11:31 AM, John Blythe <jo...@curvolabs.com> wrote:
> > Looks like part of our nightly processing was restarting the solr server
> > before all indexing was done bc of using a blunt object approach of doing
> > so at designated times, doh!
> >
> > On Tue, Aug 8, 2017 at 9:35 PM John Blythe <jo...@curvolabs.com> wrote:
> >
> >> Thanks Erick. I don't think all of those ifs are in place. Must be
> >> something in our nightly process that is conflicting. Will dive in
> tomorrow
> >> to figure out and report back.
> >>
> >> On Tue, Aug 8, 2017 at 1:27 PM Erick Erickson <er...@gmail.com>
> >> wrote:
> >>
> >>> First, are you absolutely sure you're committing before shutting down?
> >>> Hard commit in this case, openSearcher shouldn't matter.
> >>>
> >>> SolrCloud? And if not SolrCloud, how are you shutting Solr down? "Kill
> >>> -9" is evil.
> >>>
> >>> If you have transaction logs enabled then you shouldn't be losing
> >>> docs, any uncommitted docs should be replayed from the transaction
> >>> log.
> >>>
> >>> This should absolutely _not_ be happening assuming hard commits
> >>> happen. One possible explanation is that you have hard commits turned
> >>> off and are using soft commits _and_ the transaction log isn't
> >>> enabled. And you kill Solr with evil intent. But that's a long chain
> >>> of "ifs"....
> >>>
> >>> Best,
> >>> Erick
> >>>
> >>> On Tue, Aug 8, 2017 at 6:02 AM, John Blythe <jo...@curvolabs.com>
> wrote:
> >>> > hi all.
> >>> >
> >>> > i have a core that contains about 22 million documents. when the solr
> >>> > server is restarted it drops to 200-400k. the dashbaord says that
> it's
> >>> both
> >>> > optimized and current.
> >>> >
> >>> > is there config issues i need to address in solr or the server? not
> >>> really
> >>> > sure where to begin in hunting this down.
> >>> >
> >>> > thanks-
> >>>
> >> --
> >> --
> >> *John Blythe*
> >> Product Manager & Lead Developer
> >>
> >> 251.605.3071 | john@curvolabs.com
> >> www.curvolabs.com
> >>
> >> 58 Adams Ave
> >> Evansville, IN 47713
> >>
> > --
> > --
> > *John Blythe*
> > Product Manager & Lead Developer
> >
> > 251.605.3071 | john@curvolabs.com
> > www.curvolabs.com
> >
> > 58 Adams Ave
> > Evansville, IN 47713
>

Re: missing documents after restart

Posted by Erick Erickson <er...@gmail.com>.
Thanks for closing this out, I was breaking out in hives ;)

Erick

On Fri, Aug 11, 2017 at 11:31 AM, John Blythe <jo...@curvolabs.com> wrote:
> Looks like part of our nightly processing was restarting the solr server
> before all indexing was done bc of using a blunt object approach of doing
> so at designated times, doh!
>
> On Tue, Aug 8, 2017 at 9:35 PM John Blythe <jo...@curvolabs.com> wrote:
>
>> Thanks Erick. I don't think all of those ifs are in place. Must be
>> something in our nightly process that is conflicting. Will dive in tomorrow
>> to figure out and report back.
>>
>> On Tue, Aug 8, 2017 at 1:27 PM Erick Erickson <er...@gmail.com>
>> wrote:
>>
>>> First, are you absolutely sure you're committing before shutting down?
>>> Hard commit in this case, openSearcher shouldn't matter.
>>>
>>> SolrCloud? And if not SolrCloud, how are you shutting Solr down? "Kill
>>> -9" is evil.
>>>
>>> If you have transaction logs enabled then you shouldn't be losing
>>> docs, any uncommitted docs should be replayed from the transaction
>>> log.
>>>
>>> This should absolutely _not_ be happening assuming hard commits
>>> happen. One possible explanation is that you have hard commits turned
>>> off and are using soft commits _and_ the transaction log isn't
>>> enabled. And you kill Solr with evil intent. But that's a long chain
>>> of "ifs"....
>>>
>>> Best,
>>> Erick
>>>
>>> On Tue, Aug 8, 2017 at 6:02 AM, John Blythe <jo...@curvolabs.com> wrote:
>>> > hi all.
>>> >
>>> > i have a core that contains about 22 million documents. when the solr
>>> > server is restarted it drops to 200-400k. the dashbaord says that it's
>>> both
>>> > optimized and current.
>>> >
>>> > is there config issues i need to address in solr or the server? not
>>> really
>>> > sure where to begin in hunting this down.
>>> >
>>> > thanks-
>>>
>> --
>> --
>> *John Blythe*
>> Product Manager & Lead Developer
>>
>> 251.605.3071 | john@curvolabs.com
>> www.curvolabs.com
>>
>> 58 Adams Ave
>> Evansville, IN 47713
>>
> --
> --
> *John Blythe*
> Product Manager & Lead Developer
>
> 251.605.3071 | john@curvolabs.com
> www.curvolabs.com
>
> 58 Adams Ave
> Evansville, IN 47713

Re: missing documents after restart

Posted by John Blythe <jo...@curvolabs.com>.
Looks like part of our nightly processing was restarting the solr server
before all indexing was done bc of using a blunt object approach of doing
so at designated times, doh!

On Tue, Aug 8, 2017 at 9:35 PM John Blythe <jo...@curvolabs.com> wrote:

> Thanks Erick. I don't think all of those ifs are in place. Must be
> something in our nightly process that is conflicting. Will dive in tomorrow
> to figure out and report back.
>
> On Tue, Aug 8, 2017 at 1:27 PM Erick Erickson <er...@gmail.com>
> wrote:
>
>> First, are you absolutely sure you're committing before shutting down?
>> Hard commit in this case, openSearcher shouldn't matter.
>>
>> SolrCloud? And if not SolrCloud, how are you shutting Solr down? "Kill
>> -9" is evil.
>>
>> If you have transaction logs enabled then you shouldn't be losing
>> docs, any uncommitted docs should be replayed from the transaction
>> log.
>>
>> This should absolutely _not_ be happening assuming hard commits
>> happen. One possible explanation is that you have hard commits turned
>> off and are using soft commits _and_ the transaction log isn't
>> enabled. And you kill Solr with evil intent. But that's a long chain
>> of "ifs"....
>>
>> Best,
>> Erick
>>
>> On Tue, Aug 8, 2017 at 6:02 AM, John Blythe <jo...@curvolabs.com> wrote:
>> > hi all.
>> >
>> > i have a core that contains about 22 million documents. when the solr
>> > server is restarted it drops to 200-400k. the dashbaord says that it's
>> both
>> > optimized and current.
>> >
>> > is there config issues i need to address in solr or the server? not
>> really
>> > sure where to begin in hunting this down.
>> >
>> > thanks-
>>
> --
> --
> *John Blythe*
> Product Manager & Lead Developer
>
> 251.605.3071 | john@curvolabs.com
> www.curvolabs.com
>
> 58 Adams Ave
> Evansville, IN 47713
>
-- 
-- 
*John Blythe*
Product Manager & Lead Developer

251.605.3071 | john@curvolabs.com
www.curvolabs.com

58 Adams Ave
Evansville, IN 47713

Re: missing documents after restart

Posted by John Blythe <jo...@curvolabs.com>.
Thanks Erick. I don't think all of those ifs are in place. Must be
something in our nightly process that is conflicting. Will dive in tomorrow
to figure out and report back.

On Tue, Aug 8, 2017 at 1:27 PM Erick Erickson <er...@gmail.com>
wrote:

> First, are you absolutely sure you're committing before shutting down?
> Hard commit in this case, openSearcher shouldn't matter.
>
> SolrCloud? And if not SolrCloud, how are you shutting Solr down? "Kill
> -9" is evil.
>
> If you have transaction logs enabled then you shouldn't be losing
> docs, any uncommitted docs should be replayed from the transaction
> log.
>
> This should absolutely _not_ be happening assuming hard commits
> happen. One possible explanation is that you have hard commits turned
> off and are using soft commits _and_ the transaction log isn't
> enabled. And you kill Solr with evil intent. But that's a long chain
> of "ifs"....
>
> Best,
> Erick
>
> On Tue, Aug 8, 2017 at 6:02 AM, John Blythe <jo...@curvolabs.com> wrote:
> > hi all.
> >
> > i have a core that contains about 22 million documents. when the solr
> > server is restarted it drops to 200-400k. the dashbaord says that it's
> both
> > optimized and current.
> >
> > is there config issues i need to address in solr or the server? not
> really
> > sure where to begin in hunting this down.
> >
> > thanks-
>
-- 
-- 
*John Blythe*
Product Manager & Lead Developer

251.605.3071 | john@curvolabs.com
www.curvolabs.com

58 Adams Ave
Evansville, IN 47713

Re: missing documents after restart

Posted by Erick Erickson <er...@gmail.com>.
First, are you absolutely sure you're committing before shutting down?
Hard commit in this case, openSearcher shouldn't matter.

SolrCloud? And if not SolrCloud, how are you shutting Solr down? "Kill
-9" is evil.

If you have transaction logs enabled then you shouldn't be losing
docs, any uncommitted docs should be replayed from the transaction
log.

This should absolutely _not_ be happening assuming hard commits
happen. One possible explanation is that you have hard commits turned
off and are using soft commits _and_ the transaction log isn't
enabled. And you kill Solr with evil intent. But that's a long chain
of "ifs"....

Best,
Erick

On Tue, Aug 8, 2017 at 6:02 AM, John Blythe <jo...@curvolabs.com> wrote:
> hi all.
>
> i have a core that contains about 22 million documents. when the solr
> server is restarted it drops to 200-400k. the dashbaord says that it's both
> optimized and current.
>
> is there config issues i need to address in solr or the server? not really
> sure where to begin in hunting this down.
>
> thanks-