[Xapian-discuss] tcp deamon and remote connection

[Sergey Kozhukhov]

Thank you for the response. But we still can't resolve this issue. We
even make the latest stable version of xapian-core and
xapian-bindings.

On Sun, Jul 26, 2009 at 8:46 PM, Olly Betts<olly at survex.com> wrote:
> On Thu, Jul 23, 2009 at 05:19:17PM +0300, Sergey Kozhukhov wrote:
>> I'm trying to use xapian from php script with TCP daemon.
>>
>> I started TCP daemon: xapian-tcpsrv --port 5050 --writable /path-to-db
>>
>> When I trying to get access to it within my php-script with
>> remote_open_writable("localhost", 5050) or remote_open("localhost",
>> 5050) apache fails with signal 11:
>>
>> dev kernel: pid 35923 (httpd), uid 80: exited on signal 11
>
> 11 is "Segmentation fault" on Linux.  I think it is on other Unix
> platforms too, but you can probably check with "kill -l".
>
> I can't see why you should get that though.  Perhaps if you can post
> the smallest PHP test script which fails?

Actually the test script which fails is one line:
remote_open_writable("localhost", 5050);
or
remote_open("localhost", 5050);

I have also tried to run it through command line but got the same
errors: NetworkError: Closing connection.

So, even without apache it's closing connection.

Any suggestions?

>
>> TCP demon also shows me different errors:
>> Connection from 127.0.0.1, port 62652
>> Got exception NetworkError: write failed (context: /path-to-db) (Broken pipe)
>> Closing connection.
>> Connection from 127.0.0.1, port 45245
>> Got exception NetworkError: read failed (context: /path-to-db)
>> (Connection reset by peer)
>> Closing connection.
>> Connection from 127.0.0.1, port 44209
>> Got exception NetworkError: Received EOF (context: /path-to-db)
>> Closing connection.
>
> These suggest that the client is dying at different points
> mid-conversation.
>
> Actually, one possible cause of random segmentation faults is dodgy RAM
> - it might be worth running a RAM tester to check for that possibility.
> I've used memtest86+ before:
>
> http://www.memtest.org/
>
>> There are also another question: How it's better to organize xapian
>> search while using cluster of 5 servers?
>
> It depends what you're scaling for.  If each server is capable of
> servicing single queries fast enough and you're adding more servers to
> cope with the query load and provide redundancy, you can just give each
> server a copy of the database.  1.1.x provides a new database
> replication feature which helps such setups.  The documentation for
> it also describes other possible approaches, and you could use those
> with 1.0.x:
>
> http://trac.xapian.org/browser/trunk/xapian-core/docs/replication.rst
>
> If you're scaling to try to reduce the time a single query takes, then
> you could give each server a subsection of the data and combine searches
> over multiple remote databases.
>
> Cheers,
>    Olly
>

The replication for read access is not a problem for us. The problem
is how to organize one entrance for write queries? Cause each server
has it's own replicated code.



-- 
  Best Regards,
  Sergey Kozhukhov

  www.seregka.com
  E-Mail: seregka at gmail.com
  ICQ: 316817710
  Yahoo ID: kozhukhovs
  Skype: kozhukhov