Load balancing algorithm

[Dewangga Bachrul Alam]

Hello!

On 08/01/2016 04:19 PM, Andrew Hutchings wrote:
> Hi Dewangga,
> 
> I'm not quite sure what your desired outcome would be by connecting to
> two servers at the same time for a single client but it won't work the
> way you might think it will.

My main goal is, want to build real-time backup mysql server, so the
client will be sql insert to different machine at the same time. Got it?
So the scenario will be like this, if anything goes wrong at the
cluster, I can check it in cluster4-node1 and pull the data then restore
to the cluster.

> 
> I think what you might be looking for is the 'backup' keyword. With this
> NGINX will only fall back to that server if the primary servers are
> unavailable.
> 
> Using 'backup' is typically recommended for a Galera cluster anyway to
> avoid deadlocks at the commit sync due to the same table in multiple
> servers being written to at the same time. See this blog post for more
> information:
> 
> https://www.nginx.com/blog/advanced-mysql-load-balancing-with-nginx-plus/



> 
> Kind Regards
> Andrew
> 
> On 01/08/16 09:20, Dewangga Bachrul Alam wrote:
>> Hello!
>>
>> I got curios with load balancing algorithm, I got scenarios like this.
>>
>> I have 3 galera cluster, each cluster have 3 node and it was solved with
>> stream module.
>>
>> Cluster1:
>> Node1-Cluster1: 192.168.11.1
>> Node2-Cluster1: 192.168.11.2
>> Node3-Cluster1: 192.168.11.3
>>
>> Cluster2:
>> Node1-Cluster2: 192.168.12.1
>> Node2-Cluster2: 192.168.12.2
>> Node3-Cluster2: 192.168.12.3
>>
>> Cluster3:
>> Node1-Cluster3: 192.168.13.1
>> Node2-Cluster3: 192.168.13.2
>> Node3-Cluster3: 192.168.13.3
>>
>> Cluster4:
>> Node1-Cluster1: 192.168.11.1
>> Node1-Cluster4: 192.168.14.1
>>
>> But, I want to build another cluster and wont sync with galera cluster.
>> Is there any alternative instead of roundrobind, least_connected and
>> ip_hash?
>>
>> I want the algorithm is connect to backend at the sametime, looks like:
>>
>> upstream Cluster4{
>>     server 192.168.11.1:3306 max_fails=0;
>>     server 192.168.14.1:3306 max_fails=0;
>> }
>>
>> Since it's not synced, the query to upstream "Cluster4" should be not
>> round robin or asynchronous to backend, it should be simultaneously.
>>
>> Is it possible? If yes, which method should be picked up?
>>
>> _______________________________________________
>> nginx mailing list
>> nginx at nginx.org
>> http://mailman.nginx.org/mailman/listinfo/nginx
>>
>