bq: But after some updates, the index becomes unstable.
Exactly. And this kind of thing will continue to haunt you.
You said "would write to the common
directory ( but not both simultaneously for sure)". Well, how
did you guarantee that? More importantly how do you guarantee
that, for instance, both don't open a writer (whether or not
you're actually sending updates to both)? It's simply not worth
the risk IMO. Someone, somewhere will not know about the rule.
Or you'll use the wrong URL sometime. Or.... And BAM, no more
index.
Why not just go to a simple SolrCloud? All this stuff gets handled _for_ you
and you don't have to think about it.
Best,
Erick
On Fri, Jun 27, 2014 at 12:08 AM, Prasi S <prasi1...@gmail.com> wrote:
> My scenario was to have a single master and two slaves replicating from it.
> In case the Master fails, we had a failover master. But synchornizing both
> Masters was a problem. So, we planned to have a common index shared with 2
> Masters. Indexing will happen via Master1 and if it fails then it would be
> via Master2.
>
> This was not as simple as we thought. there were lot of locking issues. so
> i changed the lock type to "none" and it worked. But after some updates,
> the index becomes unstable.
>
> Thanks,
> Prasi
>
>
> On Fri, Jun 27, 2014 at 6:10 AM, Erick Erickson <erickerick...@gmail.com>
> wrote:
>
>> bq: Avoid all the ridiculous complexity of "cloud"
>>
>> And then re-introduce a single point of failure. Bad disk ==
>> unfortunate consequences....
>>
>> But frankly I don't see why you would ever _need_ to write from two
>> Solr instances. Wouldn't simply having one writer (which you could
>> change when you restarted) and multiple readers (which is OOB
>> functionality now) work?
>>
>> And as far as offloading analysis, the preanalyzed field types seem
>> interesting (although I haven't played with them)
>>
>> Erick
>>
>> On Thu, Jun 26, 2014 at 12:49 PM, Walter Underwood
>> <wun...@wunderwood.org> wrote:
>> > Cool? More like "generally useless". --wunder
>> >
>> > On Jun 26, 2014, at 12:44 PM, "Jack Krupansky" <j...@basetechnology.com>
>> wrote:
>> >
>> >> Erick, I agree, but... wouldn't it be SO COOL if it did work! Avoid all
>> the ridiculous complexity of "cloud".
>> >>
>> >> Have a temporary lock to permit and exclude updates.
>> >>
>> >> -- Jack Krupansky
>> >>
>> >> -----Original Message----- From: Erick Erickson
>> >> Sent: Thursday, June 26, 2014 12:37 PM
>> >> To: solr-user@lucene.apache.org
>> >> Subject: Re: Two solr instances access common index
>> >>
>> >> bq: But my scenario is that both solr instances would write to the
>> common
>> >> directory
>> >>
>> >> Do NOT do this. Don't even try. I guarantee Bad Things Will Happen.
>> >>
>> >> Why do you want to do this? To save disk space? Accomplish NRT
>> >> searching on multiple machines?
>> >>
>> >> Please define the problem you're trying to solve and why existing
>> supported
>> >> ways of using Solr wouldn't work for you, e.g. SolrCloud or master/slave
>> >> setups before asking for a specific solution, as this sounds very much
>> like an
>> >> XY problem.
>> >>
>> >> Best,
>> >> Erick
>> >>
>> >> On Thu, Jun 26, 2014 at 4:25 AM, Prasi S <prasi1...@gmail.com> wrote:
>> >>> Can you please tell me whihc solr version you have tried with? I tried
>> >>> giving
>> >>>
>> >>> <lockType>${solr.lock.type:none}</lockType> in 2 solr instances and
>> now it
>> >>> is working. I am not getting the write lock exception when starting the
>> >>> second instance.
>> >>>
>> >>> But my scenario is that both solr instances would write to the common
>> >>> directory ( but not both simultaneously for sure). Is there any
>> drawback of
>> >>> using "noLock"
>> >>>
>> >>> Please advice.
>> >>>
>> >>> Thanks,
>> >>> Prasi
>> >>>
>> >>>
>> >>> On Thu, Jun 26, 2014 at 3:20 PM, Uwe Reh <r...@hebis.uni-frankfurt.de>
>> wrote:
>> >>>
>> >>>> Hi,
>> >>>>
>> >>>> with the lock type 'simple' I have tree instances (different JREs,
>> >>>> GC-Problem) running on the same files.
>> >>>> You should use this option only for a readonly system. Otherwise it's
>> easy
>> >>>> to corrupt the index.
>> >>>>
>> >>>> Maybe you should have a look on replication or SolrCloud.
>> >>>>
>> >>>> Uwe
>> >>>>
>> >>>>
>> >>>> Am 26.06.2014 11:25, schrieb Prasi S:
>> >>>>
>> >>>> Hi,
>> >>>>> Is it possible to point two solr instances to point to a common index
>> >>>>> directory. Will this work wit changing the lock type?
>> >>>>>
>> >>>>>
>> >>>>>
>> >>>>> Thanks,
>> >>>>> Prasi
>> >>>>>
>> >>
>> >
>> > --
>> > Walter Underwood
>> > wun...@wunderwood.org
>> >
>> >
>> >
>>
