SphinxSE on MySQL 5.1.42 and Sphinx 0.9.9

Is there anyone, who has successfully compilled SphinxSE using MySQL 5.1.42 and Sphinx 0.9.9 sources?

Sphinx forum is completely quiet about this…

[URL]sphinx engine 0.9.9 compilation problem with xtradb 5.1.42 | Sphinx

…and sphinx docs for Sphinx SE haven’t been updated for a long time

As I remember, Vadim wrote the storage engine for 5.0 and it has not been ported to 5.1. I’ll ask him to post a reply here, but I think it needs to be ported to 5.1, which we would look for sponsorship to do. The 5.1 storage engine API is not fully compatible with the 5.0 API.

Vadim corrected me, there is a 5.1 plugin storage engine. But he says he hasn’t tested it for years.

Obviously I am not an expert in this topic. The guys who are experts in it are very busy and are not very active on this forum. But if you can’t solve the problem, place an inquiry on percona.com and we can give you an estimate of what it would take for us to solve this.

Finally, our sysadmin managed to compile it. Thanks to Andrew Garner at sphinx forum:

[URL]sphinx engine 0.9.9 compilation problem with xtradb 5.1.42 | Sphinx

Great )

I have the same problem with SphinxSE 0.9.9 and MYSQL 5.1.41.
It’s the default version of MYSQL on Ubuntu 10.04 LTS.

It compiles fine and works, although this message is always in the error log “[Warning] Storage engine ‘SPHINX’ has conflicting typecode. Assigning value 42.” and queries through the plugin fail under high load, when that happens connection gets dropped and CPU usage goes to 100% on MySQL.

Can someone help me fix this Sphinx problem?

[B]andgus wrote on Tue, 04 May 2010 14:33[/B]
this message is always in the error log "[B][Warning] Storage engine 'SPHINX' has conflicting typecode. Assigning value 42.[/B]"

I get the same message after mysqld startup, but it doesn’t seem to be a problem.

[B]andgus wrote on Tue, 04 May 2010 14:33[/B]

queries through the plugin fail under high load, when that happens connection gets dropped and CPU usage goes to 100% on MySQL.

I have not seen this behaviour yet, are you sure it is query-type independent? Is there anything in the searchd log?

Andgus,

I have tested the scalability on a dual-core cpu and the behaviour is bad - the plugin has negative scalability.

1 thread:
Requests per second: 5.29 [#/sec] (mean)

2 threads:
Requests per second: 6.43 [#/sec] (mean)

4 threads:
Requests per second: 6.85 [#/sec] (mean)

8 threads:
Requests per second: 6.41 [#/sec] (mean)

16 threads:
Requests per second: 4.01 [#/sec] (mean)

32 threads:
Requests per second: 3.63 [#/sec] (mean)

These results above are from apachebench calling php script, thus if you use direct connections to mysqld, the result might be even worse.

When I run these queries to the searchd directly, the scalability is OK.