[Pgpool-general] Help in pgpool loadbalancing

Nimesh Satam nimesh.satam at gmail.com
Mon Jul 20 11:19:38 UTC 2009


Glyn/Steven,


Thanks for your reply.

What we are trying to do here is: when we are doing a database vacuum we
don't want any connections for the given node. It might be any of the nodes
in the replication.

Over the weekend when we tried to change the weight of the backend-1 to 0
and reload using the pgpool reload facility the node was still receiving
connections. Hence I am not sure if what I am trying to do is correct or
not.

Can you tell me how I can achieve this? I just want to make sure that we do
not send any Select queries to a certain node for a certain period of time.

Regards,
Nimesh.

On Mon, Jul 20, 2009 at 1:33 AM, Steven Crandell
<steven.crandell at gmail.com>wrote:

> As far as I understand things, you would still be seeing
> inserts/updates/deletes on backend node 0 even if it is weighted to 0
> because pgpool considers this node to be the master and therefore the only
> node capable of accepting new data.
> e.g. the weight of a node in a master/slave configuration only determines
> the number of selects being sent to slaves.
>
> I run a 3 node cluster with mammoth replicator replicating data from master
> to slaves and have set up pgpool with node 0 set to weight 0 and the other
> two slaves set to 0.5.
>
> My intention with this configuration is for the master to only ever receive
> INSERT/UPDATE/DELETE statement and to have all SELECT statements sent to the
> two slaves in a 50%/50% split.  Thereby, the cluster is able to scale up to
> the point where the number of writes exceed the capabilities of the master
> without wasting master system resources on reads.
>
> So far this setup seems to work but I must admit that this configuration
> does not seem to be clearly documented anywhere that I could find despite it
> being a fairly typical need for a 3rd party replication solutions.
>
> If my assumptions about this are wrong I hope someone with more experience
> in this area will pipe up.
>
> -s
>
> On Sun, Jul 19, 2009 at 11:29 AM, Glyn Astill <glynastill at yahoo.co.uk>wrote:
>
>>
>> > From: Nimesh Satam <nimesh.satam at gmail.com>
>> > Subject: [Pgpool-general] Help in pgpool loadbalancing
>> > To: pgpool-general at pgfoundry.org
>> > Date: Sunday, 19 July, 2009, 4:04 PM
>> > Hi,
>> >
>> > Can anybody highlight how queries can be loadbalanced
>> > in pgpool? How does backend_weight0 means and what happens
>> > if the value for this parameter is set to '0'.
>> >
>> > I need to use this parameter in the given below way:
>> >
>> >
>> > Set backend_weight0 = 0
>> >
>> > So as no queries go to the first node. But in our
>> > usage we do see queries coming to the node for whom the
>> > parameter is set to 0. We are using pgpool in master/slave
>> > mode, with no connection caching.
>> >
>> >
>>
>> If you do not want any queries to go through to the first node, why not
>> just remove it completely from the config?
>>
>>
>>
>> _______________________________________________
>> Pgpool-general mailing list
>> Pgpool-general at pgfoundry.org
>> http://pgfoundry.org/mailman/listinfo/pgpool-general
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://pgfoundry.org/pipermail/pgpool-general/attachments/20090720/95733004/attachment-0001.html>


More information about the Pgpool-general mailing list