Hello Serafin,
thanks for sharing your experiences.
On 2013-06-06 22:37, Serafin wrote:
I've just come across this thread, so my personal experience
I've implemented MailPiler for my customer a few monts ago. The mail
throughput for the server is about 5k mails a day.
Also, the old archive was imported there. Old archive was historical
MailArchiva which occupied about 1T when we started import. After
import
to MailPiler archive occupies less than 500GB
was it the open source edition of mailarchiva or the commercial
version?
In the matter of performance - archiva, MySQL db and Sphinx are on
separate Luns connected via FC. Machine has 4 CPUs and 8GB RAM and it
is
Ubuntu 12.04 hosted on vSphere.
I don't know much more details about hardware, as this is customer's
environment.
In the matter of performance, initially we had some issues with
Sphinx,
but problem was very simple - delta indexing during import was taking
so
long, that next iteration of index rotate has been started before
previous one was finished.
Increasing time between iterations and some basic tuning of MySQL and
Sphinx solved the problem.
that's indeed a solution, however in the meantime a perhaps more
elegant
one popped up: use several main index options in sphinx.conf:
source main1 : base
{
sql_query_pre = SET NAMES utf8
sql_query = SELECT id, `from`, `to`, `fromdomain`, `todomain`,
`subject`, `arrived`, `sent`, `body`, `size`, `direction`, `folder`,
`attachments`, `attachment_types` FROM sph_index WHERE id=-1;
}
source main2 : base
{
sql_query_pre = SET NAMES utf8
sql_query = SELECT id, `from`, `to`, `fromdomain`, `todomain`,
`subject`, `arrived`, `sent`, `body`, `size`, `direction`, `folder`,
`attachments`, `attachment_types` FROM sph_index WHERE id=-1;
}
source main3 : base
{
sql_query_pre = SET NAMES utf8
sql_query = SELECT id, `from`, `to`, `fromdomain`, `todomain`,
`subject`, `arrived`, `sent`, `body`, `size`, `direction`, `folder`,
`attachments`, `attachment_types` FROM sph_index WHERE id=-1;
}
source main4 : base
{
sql_query_pre = SET NAMES utf8
sql_query = SELECT id, `from`, `to`, `fromdomain`, `todomain`,
`subject`, `arrived`, `sent`, `body`, `size`, `direction`, `folder`,
`attachments`, `attachment_types` FROM sph_index WHERE id=-1;
}
and finally in config.php:
$config['SPHINX_MAIN_INDEX'] = 'main1,main2,main3,main4';
I'll make this the default in the shipped config.php of the gui. Then
if the main1 index files grow
large, then just replace main1 with main2 in the cron jobs. Again later
if main2 grows large, too, you may switch to main3, etc.
I have no contact with the customer since February, so I assume
MailPiler works perfectly and without issues
I'm glad to hear that.
Janos
