| 1 |
Benjamin MALYNOVYTCH wrote: |
| 2 |
> Hi ! |
| 3 |
> |
| 4 |
> I' would like to know if using high availability cluster (with master |
| 5 |
> and slave), as described in |
| 6 |
> http://www.gentoo.org/proj/en/cluster/hpc-howto.xml would allow to |
| 7 |
> have also load balancing ? |
| 8 |
> Will the slave take a part of the load of the master when both of the |
| 9 |
> servers are up and available ? Or will it be waiting the |
| 10 |
> unavailability to work ? |
| 11 |
|
| 12 |
You have almost two ways: |
| 13 |
- you can add a load balancer in front of your cluster (you can |
| 14 |
implement it with keepalived) |
| 15 |
- you can distribute active services on the nodes (two or more) |
| 16 |
for example: |
| 17 |
|
| 18 |
if you have 2 nodes acting as a cluster of web/mail/DB servers, you can |
| 19 |
setup a DRBD daemon with 2 or 3 resources, add the services to the nodes |
| 20 |
and use heartbeat to distribute active services on the cluster. |
| 21 |
For example you can have apache active on all the nodes, postfix active |
| 22 |
on the first node and mysql active on the second node. |
| 23 |
In case of failure all the services will became active on the remaining |
| 24 |
node. |
| 25 |
|
| 26 |
In the future, with drbd 8.0 and GFS you can improve this kind of setup. |
| 27 |
|
| 28 |
The solution depends on you budget, you can also implement it with an |
| 29 |
ata over ethernet disk array and a couple of servers. |
| 30 |
|
| 31 |
-- |
| 32 |
Christian Zoffoli : Gentoo Developer |
| 33 |
Mail : xmerlin at gentoo.org |
| 34 |
Web : http://dev.gentoo.org/~xmerlin |
| 35 |
-- |
| 36 |
gentoo-cluster@g.o mailing list |
Archives