Hi Ralph, As of my experience, the cluster stretched between 2 DCs is a good idea only if: 1. You have either a super reliable network, like 3 independent links from different providers, or you have the majority of nodes in one DC, let's say 2/3 nodes are located in DC-1 and 1/3 in DC-2 2. Your interconnection speed is at least 10G between DCs and really small latency, like 3-5 ms (500Mbps and 1G is too slow for the production with high workload, but enough for POC) To achieve more data durability and availability - and this is the cornerstone of your idea I believe, - I would suggest implementing rbd mirroring for block storage and multizone deployment for object-store. With regards, Vladimir. On Tue, Aug 10, 2021 at 3:58 PM Ralph Soika <ralph.soika@xxxxxxxxx> wrote: > Hi, > > I have a question about architecture: > > Normally it is recommended to build a Ceph Cluster within one Data > center. In a Data center I have a network with 1Gbit/s and it is working > great with my Kubernetes Cluster running in the same data center. > > For more reliability I plan to add ceph nodes running in a different > data center. The data centers are not connected directly and need to > communicate via public IPs. I did some tests and - not surprising - the > network performance is going down to 500Mbit/s between two nodes in two > different data centers. > > My questions are: > > 1. Is 500Mbit/s to slow to run a ceph cluster? > 2. Is it a bad idea to build a Ceph Cluster with nodes in different > data centers if the network bandwith is only 1Gbit/s? > > The background: My idea was that the cluster becomes more robust because > I am no longer bound to a single data center (which also implies a > vendor lock-in). > > What is your opinion about this idea running nodes in different data > centers? > > > Thanks in advance > > ==== > Ralph > _______________________________________________ > ceph-users mailing list -- ceph-users@xxxxxxx > To unsubscribe send an email to ceph-users-leave@xxxxxxx > _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |