Hi, We use the 389 Directory Server version 1.4.2.15. In the documentation of the Red Hat Directory Server it says, as many as 20 masters are supported in an MMR. It sounds to be a hardcoded limitation defined to avoid overloaded servers and network. Shouldn't it be depending on the MMR topology, i.e. rather on the total number of replication agreements within the whole scenario? Or does "20 masters" indeed mean that the limitation of the MMR topology is the number of 380 replication agreements as shown in the "fully connected mesh" scenario (https://access.redhat.com/documentation/en-us/red_hat_directory_server/10/ht...) with 20 masters and 20x19 replication agreements? Is there someone who possibly has experience with scenarios of more than 20 master servers? Thanks, Eugen
Hi Eugen,
I think that is what was tested by Red Hat and not necessarily a hard limit.
Regards,
Paul M. Whitney paul.whitney@mac.com
Sent from my Mac Book Pro
On Sep 30, 2020, at 9:56 AM, Eugen Lamers eugen.lamers@br-automation.com wrote:
Hi, We use the 389 Directory Server version 1.4.2.15. In the documentation of the Red Hat Directory Server it says, as many as 20 masters are supported in an MMR. It sounds to be a hardcoded limitation defined to avoid overloaded servers and network. Shouldn't it be depending on the MMR topology, i.e. rather on the total number of replication agreements within the whole scenario? Or does "20 masters" indeed mean that the limitation of the MMR topology is the number of 380 replication agreements as shown in the "fully connected mesh" scenario (https://access.redhat.com/documentation/en-us/red_hat_directory_server/10/ht...) with 20 masters and 20x19 replication agreements? Is there someone who possibly has experience with scenarios of more than 20 master servers? Thanks, Eugen _______________________________________________ 389-users mailing list -- 389-users@lists.fedoraproject.org To unsubscribe send an email to 389-users-leave@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/389-users@lists.fedoraproject....
On 1 Oct 2020, at 05:25, Paul Whitney paul.whitney@mac.com wrote:
Hi Eugen,
I think that is what was tested by Red Hat and not necessarily a hard limit.
Correct, however Red Hat also tested up to 60 for FreeIPA. Certainly I know there is a deployment in the world that has scaled up past 1000 servers.
Generally once you get past say 8 servers, you need to think about your replication topology and how the data will flow. The largest sites I know tend to go to maximum 8 write-accepting servers, and then replicate to N hubs and replicas that are read only past that for the best performance and reliability.
Regards,
Paul M. Whitney paul.whitney@mac.com
Sent from my Mac Book Pro
On Sep 30, 2020, at 9:56 AM, Eugen Lamers eugen.lamers@br-automation.com wrote:
Hi, We use the 389 Directory Server version 1.4.2.15. In the documentation of the Red Hat Directory Server it says, as many as 20 masters are supported in an MMR. It sounds to be a hardcoded limitation defined to avoid overloaded servers and network. Shouldn't it be depending on the MMR topology, i.e. rather on the total number of replication agreements within the whole scenario? Or does "20 masters" indeed mean that the limitation of the MMR topology is the number of 380 replication agreements as shown in the "fully connected mesh" scenario (https://access.redhat.com/documentation/en-us/red_hat_directory_server/10/ht...) with 20 masters and 20x19 replication agreements? Is there someone who possibly has experience with scenarios of more than 20 master servers? Thanks, Eugen _______________________________________________ 389-users mailing list -- 389-users@lists.fedoraproject.org To unsubscribe send an email to 389-users-leave@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/389-users@lists.fedoraproject....
389-users mailing list -- 389-users@lists.fedoraproject.org To unsubscribe send an email to 389-users-leave@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/389-users@lists.fedoraproject....
— Sincerely,
William Brown
Senior Software Engineer, 389 Directory Server SUSE Labs, Australia
On 10/1/20 1:05 AM, William Brown wrote:
On 1 Oct 2020, at 05:25, Paul Whitney paul.whitney@mac.com wrote:
Hi Eugen,
I think that is what was tested by Red Hat and not necessarily a hard limit.
Correct, however Red Hat also tested up to 60 for FreeIPA. Certainly I know there is a deployment in the world that has scaled up past 1000 servers.
Generally once you get past say 8 servers, you need to think about your replication topology and how the data will flow. The largest sites I know tend to go to maximum 8 write-accepting servers, and then replicate to N hubs and replicas that are read only past that for the best performance and reliability.
That is right, there is no hard limit regarding the number of masters. Even the number of hosts (master,hub,consumer) in the topology is not limited. Freeipa tested with up to 60 masters this is why there are some recommendations in the doc. Some considerations with a high number of hosts is the #replication agreement and the max hop to replicate an update from/to hosts. #replication agreement will create "pressure" on the changelog and could compete with changelog updates (can trigger retry). #hop will create latency. Some common deployments are 4-8 masters, 4-8 hubs and many consumers.
regards thierry
Regards,
Paul M. Whitney paul.whitney@mac.com
Sent from my Mac Book Pro
On Sep 30, 2020, at 9:56 AM, Eugen Lamers eugen.lamers@br-automation.com wrote:
Hi, We use the 389 Directory Server version 1.4.2.15. In the documentation of the Red Hat Directory Server it says, as many as 20 masters are supported in an MMR. It sounds to be a hardcoded limitation defined to avoid overloaded servers and network. Shouldn't it be depending on the MMR topology, i.e. rather on the total number of replication agreements within the whole scenario? Or does "20 masters" indeed mean that the limitation of the MMR topology is the number of 380 replication agreements as shown in the "fully connected mesh" scenario (https://access.redhat.com/documentation/en-us/red_hat_directory_server/10/ht...) with 20 masters and 20x19 replication agreements? Is there someone who possibly has experience with scenarios of more than 20 master servers? Thanks, Eugen _______________________________________________ 389-users mailing list -- 389-users@lists.fedoraproject.org To unsubscribe send an email to 389-users-leave@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/389-users@lists.fedoraproject....
389-users mailing list -- 389-users@lists.fedoraproject.org To unsubscribe send an email to 389-users-leave@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/389-users@lists.fedoraproject....
— Sincerely,
William Brown
Senior Software Engineer, 389 Directory Server SUSE Labs, Australia _______________________________________________ 389-users mailing list -- 389-users@lists.fedoraproject.org To unsubscribe send an email to 389-users-leave@lists.fedoraproject.org Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/ List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines List Archives: https://lists.fedoraproject.org/archives/list/389-users@lists.fedoraproject....
Hi, many thanks for your answers. Can someone describe this test scenario with 60 masters? Is the topology available to have a look at? I think this will not be a fully meshed scenario (multi-master with meshed replication threads), right? It would be nice to have some insight into the "test" scenario: - How many replication agreements were created and how were the changelogs set up? - Are there measurement results for the utilization and data flow? Regards, Eugen
389-users@lists.fedoraproject.org