Mark R Bannister wrote:
Is it possible to configure syncrepl on the consumer side with a failover policy so that if one provider goes down it can automatically retry against another?
On each consumer you should use separate syncrepl directives for each provider you want to get data from. Two providers -> two syncrepl statements. Note the hint about RID values for each syncrepl statement in the docs.
That would remove the need for separate load balancers.
No need for any kind of load balancers serving syncrepl. I'd rather consider load balancers for replication harmful.
You may want to have a look in the admin guide:
http://www.openldap.org/doc/admin24/replication.html
Ciao, Michael.
On 13/02/2015 18:54, Michael Ströder wrote:
Mark R Bannister wrote:
Is it possible to configure syncrepl on the consumer side with a failover policy so that if one provider goes down it can automatically retry against another?
On each consumer you should use separate syncrepl directives for each provider you want to get data from. Two providers -> two syncrepl statements. Note the hint about RID values for each syncrepl statement in the docs.
That would remove the need for separate load balancers.
No need for any kind of load balancers serving syncrepl. I'd rather consider load balancers for replication harmful.
You may want to have a look in the admin guide:
http://www.openldap.org/doc/admin24/replication.html
Ciao, Michael.
Thanks Michael, I had looked at the admin guide in some depth, but obviously missed the bit that said you could have multiple syncrepl directives. So what you're saying is, in a multi-master set-up with two masters and multiple replicas, each replica would typically have two syncrepl statements, one for each of the masters? And that would mean we could take out any one master server and not affect the ability to replicate down the chain?
Thanks, Mark.
Mark R Bannister wrote:
On 13/02/2015 18:54, Michael Ströder wrote:
Mark R Bannister wrote:
Is it possible to configure syncrepl on the consumer side with a failover policy so that if one provider goes down it can automatically retry against another?
On each consumer you should use separate syncrepl directives for each provider you want to get data from. Two providers -> two syncrepl statements. Note the hint about RID values for each syncrepl statement in the docs.
That would remove the need for separate load balancers.
No need for any kind of load balancers serving syncrepl. I'd rather consider load balancers for replication harmful.
So what you're saying is, in a multi-master set-up with two masters and multiple replicas, each replica would typically have two syncrepl statements, one for each of the masters? And that would mean we could take out any one master server and not affect the ability to replicate down the chain?
Yes.
Ciao, Michael.
openldap-technical@openldap.org