[Pgpool-hackers] Admin interface, statistics views, etc

[Guillaume Lelarge]

Le 10/08/2010 17:26, Tatsuo Ishii a écrit :
>> Le 07/08/2010 01:16, Tatsuo Ishii a écrit :
>>>> Le 06/08/2010 03:14, Tatsuo Ishii a écrit :
>>>>>> Le 03/08/2010 13:29, Guillaume Lelarge a écrit :
>>>>>>> Le 03/08/2010 12:23, Tatsuo Ishii a écrit :
>>>>>>>>> Le 03/08/2010 03:48, Tatsuo Ishii a écrit :
>>>>>>>>>>>>> AFAICT, it works really good. We still miss pgsql_pid. And should I add
>>>>>>>>>>>>> some other reports? what do other people need?
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> No comments on this patch? meaning I should commit it or not?
>>>>>>>>>>>
>>>>>>>>>>> I was waiting for you add pgsql_pid. Have you done it?
>>>>>>>>>>
>>>>>>>>>> I have added new function:
>>>>>>>>>> /*
>>>>>>>>>>  * Return pointer to i th child, j th connection pool and k th backend
>>>>>>>>>>  * of connection info on shmem.
>>>>>>>>>>  */
>>>>>>>>>> ConnectionInfo *pool_coninfo(int child, int connection_pool, int backend)
>>>>>>>>>>
>>>>>>>>>> I hope this will make your implementation regarding pgsql_pid easier.
>>>>>>>>>> For example, pool_coninfo(10, 0, 1) will return pointer to
>>>>>>>>>> ConnectionInfo corresponds to 1th backend pid which is in 0th
>>>>>>>>>> connection pool in 10th pgpool child.
>>>>>>>>>
>>>>>>>>> Sorry for not answering sooner. I thought you were working on adding
>>>>>>>>> pgsql_pid. There was a little misunderstanding :)
>>>>>>>>
>>>>>>>> Oh, I thought I woould work on pcp command, you would work on show
>>>>>>>> command. 
>>>>>>>>
>>>>>>>
>>>>>>> No problem :)
>>>>>>>
>>>>>>> I'll work on it ASAP.
>>>>>>>
>>>>>>
>>>>>> With your last commits, it makes it quite easy to do. Here is my new
>>>>>> patch. I still have an issue and perhaps you have an idea on how to
>>>>>> resolve it. Is there a way to know which connection is the active one in
>>>>>> a pool?
>>>>>
>>>>> This patch looks good and works good for me. Please commit.
>>>>> Would you please update doc/pgpool-en.html as well to add those new
>>>>> show commands?
>>>>
>>>> Do I have time to commit it only this monday (2010-08-09)? I would
>>>> really like to have this in pgPool-II 3.0, but I think I won't have time
>>>> to work on this before mondy morning.
>>>
>>> No problem, please commit when you have time.
>>
>> Done.
>>
>> I had to fix my patch because I didn't take care the backend id. Right
>> now, it only looks at backend 0. Not sure this is a sensible thing to
>> do, but it would be easy to fix if it's not.
> 
> Can you be more specific?
> 

Sure. In the function processes_reporting() (file
pool_process_reporting.c), there is this code:

for (pool = 0; pool < pool_config->max_pool; pool++)
{
    poolBE = pool*MAX_NUM_BACKENDS;

pool is multiplied by MAX_NUM_BACKENDS to get the right pool id for the
backend 0. If I didn't use this operation, I would access pool of
inexisting backends.


-- 
Guillaume
 http://www.postgresql.fr
 http://dalibo.com