[pgpool-general: 3983] Re: Possible connection leak
Tatsuo Ishii
ishii at postgresql.org
Tue Aug 18 12:30:58 JST 2015
Why do you think it's a problem?
As I said, reaching to the num_init_children limit itself is not a
problem as far as there's "Connected" = 0 process exists.
Best regards,
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese:http://www.sraoss.co.jp
> When I switch back to the old database using the "\c", it creates a new process, rather than reuse the existing old one.
>
> Kind regards,
> Derek.
>
>> On 18 Aug 2015, at 12:59 pm, Tatsuo Ishii <ishii at postgresql.org> wrote:
>>
>> That's no problem. Because the old one ("Connected" = 0) can accept
>> new connection. As long as there are process which have ("Connected" =
>> 0), pgpool-II happily accepts new connections.
>>
>> You can verify it by setting num_init_children = 2 and do:
>>
>> psql db1
>> \c db2
>> \c db1
>> \c db2
>> :
>> :
>>
>> Best regards,
>> --
>> Tatsuo Ishii
>> SRA OSS, Inc. Japan
>> English: http://www.sraoss.co.jp/index_en.php
>> Japanese:http://www.sraoss.co.jp
>>
>>> The old one has a value of 0. But remains in the list until the child life time is reached.
>>>
>>> Kind regards,
>>> Derek.
>>>
>>>> On 18 Aug 2015, at 12:46 pm, Tatsuo Ishii <ishii at postgresql.org> wrote:
>>>>
>>>> Does not reporduce here. Are sure that "Connected" field is 1 for both
>>>> the new and old database?
>>>>
>>>> Database : test2
>>>> Username : t-ishii
>>>> Start time : 2015-08-18 11:40:05
>>>> Creation time: 2015-08-18 11:44:17
>>>> Major : 3
>>>> Minor : 0
>>>> Counter : 1
>>>> Backend PID : 6564
>>>> Connected : 1
>>>> PID : 6534
>>>> Backend ID : 0
>>>>
>>>> Best regards,
>>>> --
>>>> Tatsuo Ishii
>>>> SRA OSS, Inc. Japan
>>>> English: http://www.sraoss.co.jp/index_en.php
>>>> Japanese:http://www.sraoss.co.jp
>>>>
>>>>> Hi,
>>>>>
>>>>> Psql to pgpool.
>>>>>
>>>>> After logging into the Postgres database, then switch to another database, it would have 2 processes according to pcp_proc_info.
>>>>>
>>>>>
>>>>> Kind regards,
>>>>> Derek.
>>>>>
>>>>> On 18 Aug 2015, at 11:59 am, Tatsuo Ishii <ishii at postgresql.org> wrote:
>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> I'm current running pgpool-II 3.4.3-1 using pgpoolAdmin 3.4.1 on Rhel7.1.
>>>>>>>
>>>>>>> When I connect using psql to database1 via the virtual ip, then later switch over to another database2 using the "\c". Then switch back to database1. Pgpool creates 3 connections, which I can see in pgpoolAdmin->status->processes.
>>>>>>>
>>>>>>> If I continue to switch between the two databases, it will continue to increase the number of pgpool connections via the virtual ip. Eventually, these connections timeout when it reaches the child_life_time value or if it reaches the num_init_children limit, it will hang until a connection is available.
>>>>>>
>>>>>> I don't know when you say "the number of pgpool connections", these
>>>>>> are from client (psql) to pgpool or pgpool to PostgreSQL. Because the
>>>>>> former will not increase as long as the number of psql you are using,
>>>>>> I assume you are referring to connections from pgpool to
>>>>>> PostgreSQL.
>>>>>>
>>>>>> The number could increase until num_init_children reaches. However
>>>>>> some of process are expected to accept connections from clients. You
>>>>>> could verify it by using "show pool_nodes". "pool_connected" column
>>>>>> indicates that the process is ready to accept connections from clients
>>>>>> (in this case 0).
>>>>>>
>>>>>> Please remeber, however, that psql keeps old connection until \c
>>>>>> succeeds. For example, if num_init_children = 1, you are never able to
>>>>>> use \c.
>>>>>>
>>>>>>> I have the connection_cache on.
>>>>>>>
>>>>>>> Is there something that I need to set?
>>>>>>
>>>>>> You might want to set client_idle_limit.
>>>>>>
>>>>>> Best regards,
>>>>>> --
>>>>>> Tatsuo Ishii
>>>>>> SRA OSS, Inc. Japan
>>>>>> English: http://www.sraoss.co.jp/index_en.php
>>>>>> Japanese:http://www.sraoss.co.jp
More information about the pgpool-general
mailing list