Clustering question (Roller 2.0)

[Bryan Field-Elliot <br...@netmeme.org>]

We have two instances of Roller running, backed by the same database,
same file uploads directory, same Lucene search engine directory, etc. I
just wanted to double check that this should be fine, no gotchas with
file contention etc? I'm particularly curious about the Lucene search
index directory, is it going to be just fine when shared across multiple
running instances who may try to add to the index at the same time?

Re: Clustering question (Roller 2.0)

[Anil Gangolli <an...@busybuddha.org>]

 From what I could find 
(http://lucene.apache.org/java/docs/fileformats.html) Lucene uses files 
in the directory determined by java.io.tmpdir for locking on the index 
files.  Are the two instances sharing this directory?  If not, that's a 
problem.

Anil Gangolli wrote:

>
> I haven't got additional information on the search-index sharing 
> question, but to avoid duplicate weblog update pings sent by both 
> hosts, you will want to set *pings.queueProcessingIntervalMins=0* on 
> one of them.  There should be db-based synchronization for all of the 
> periodic tasks to better support clustering, but there isn't currently.
>
> --a.
>
> Dave Johnson wrote:
>
>>
>> On Dec 5, 2005, at 8:50 AM, Bryan Field-Elliot wrote:
>>
>>> We have two instances of Roller running, backed by the same database,
>>> same file uploads directory, same Lucene search engine directory, 
>>> etc. I
>>> just wanted to double check that this should be fine, no gotchas with
>>> file contention etc? I'm particularly curious about the Lucene search
>>> index directory, is it going to be just fine when shared across 
>>> multiple
>>> running instances who may try to add to the index at the same time?
>>
>>
>>
>> I don't know much about the inner workings of Lucene, but that 
>> certainly sounds like a problem.
>>
>> Any Lucene experts out there?
>>
>> - Dave
>>
>>
>
>

Re: Clustering question (Roller 2.0)

[Anil Gangolli <an...@busybuddha.org>]

I haven't got additional information on the search-index sharing 
question, but to avoid duplicate weblog update pings sent by both hosts, 
you will want to set *pings.queueProcessingIntervalMins=0* on one of 
them.  There should be db-based synchronization for all of the periodic 
tasks to better support clustering, but there isn't currently.

--a.

Dave Johnson wrote:

>
> On Dec 5, 2005, at 8:50 AM, Bryan Field-Elliot wrote:
>
>> We have two instances of Roller running, backed by the same database,
>> same file uploads directory, same Lucene search engine directory, etc. I
>> just wanted to double check that this should be fine, no gotchas with
>> file contention etc? I'm particularly curious about the Lucene search
>> index directory, is it going to be just fine when shared across multiple
>> running instances who may try to add to the index at the same time?
>
>
> I don't know much about the inner workings of Lucene, but that 
> certainly sounds like a problem.
>
> Any Lucene experts out there?
>
> - Dave
>
>

Re: Clustering question (Roller 2.0)

[Dave Johnson <da...@rollerweblogger.org>]

On Dec 5, 2005, at 8:50 AM, Bryan Field-Elliot wrote:
> We have two instances of Roller running, backed by the same database,
> same file uploads directory, same Lucene search engine directory, etc. 
> I
> just wanted to double check that this should be fine, no gotchas with
> file contention etc? I'm particularly curious about the Lucene search
> index directory, is it going to be just fine when shared across 
> multiple
> running instances who may try to add to the index at the same time?

I don't know much about the inner workings of Lucene, but that 
certainly sounds like a problem.

Any Lucene experts out there?

- Dave