[pgpool-general: 4709] Re: Pgpool and postgis
paolo.crosato at gmail.com
Sat May 28 02:50:07 JST 2016
Thank you both for the answers.
I will test postgis replication across the nodes and blacklist the
functions that need writing or updating.
Both the dbs and the watchdogs reside on the same data center, so hopefully
there will be no issues with network performances. I will monitor that too
2016-05-26 14:44 GMT+02:00 Muhammad Usama <m.usama at gmail.com>:
> On Thu, May 26, 2016 at 2:29 PM, Tatsuo Ishii <ishii at postgresql.org>
>> > Hi,
>> > I'm evaluating a pgpool cluster with 2 nodes as a substitute for the
>> > postgresql instance we are using now.
>> >>From an application perspective, my main concern is postgis. We are
>> > extensively using postgis in our applications. As I understood pgpool
>> > transparently handle any pgsql write instruction like insert or update,
>> > how are statements like AddGeometryColumn handled? Do I need to put
>> > "write" statement in the black_function_list?
>> Basically yes. Recently Usama eliminated the space limitation of the
>> memory area for black_function_list and you can register as many as
>> function names you like.
>> Usama, do you have any suggestion for this?
> Not really. Basically black_function_list is the best way to make sure
> that the statement with writing function lands on the master node.
> Alternatively query can be prefixed with /*NO LOAD BALANCE*/ comments,
> but that will require the changes on the application side.
> Another thing to note here is black_function_list and whie_function_list
> also support regular expression, So if there is a specific prefix that all
> writing functions have than instead of putting the name of each and every
> function in the list a generic regular expression can also be used.
> Muhammad Usama
>> > As a side note, we have been provided with 2 db nodes with the same
>> size as
>> > our singel instance, but the two watchdog machines are 2 cores instances
>> > with only 1G or ram, is this enough to run the watchdog smoothly? We
>> > around 1K IOPS.
>> Pgpool (and watchdog) does not consume too much CPU and memory and
>> usually the resource you are given should be enough to handle the
>> load. I think network device performance is rather important, since
>> pgpool needs to handle lots of network traffics.
>> Best regards,
>> Tatsuo Ishii
>> SRA OSS, Inc. Japan
>> English: http://www.sraoss.co.jp/index_en.php
>> pgpool-general mailing list
>> pgpool-general at pgpool.net
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the pgpool-general