You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@cloudstack.apache.org by Bryan Tiang <br...@hotmail.com> on 2025/03/03 14:39:37 UTC

Re: Disaster Recovery Strategies for Cloudstack Management Server

Hey Jurgen,

Thanks for the response! Yea no Active-Active MYSQL… That’s suicide with no conflict resolutions.

How do you balance the traffic between the Geo Redundant ACS Setup? Im guessing this is only convenient if you use subdomains in your CS Agents, is that what you managed to achieve?

I’ll look into the multiple mgmt servers again, I couldn’t get it working the last time.

Regards,
Bryan
On 28 Feb 2025 at 6:51 PM +0800, Jürgen Gotteswinter <ju...@mgm-tp.com.INVALID>, wrote:
> Sorry for the confusion, 2-node would be active/passive
>
> Am 28.02.25, 11:47 schrieb "Jürgen Gotteswinter" <juergen.gotteswinter@mgm-tp.com.INVA <ma...@mgm-tp.com.INVA>LID>:
>
>
> Regarding #3
>
>
> We are running a geo redundant acs management and galera database spanning 3 datacenters. So far, no issues
>
>
> See https://docs.cloudstack.apache.org/en/latest/adminguide/reliability.html <https://docs.cloudstack.apache.org/en/latest/adminguide/reliability.html>
>
>
> Active/Active should be also doable using keepalived and standard mysql replication. But this will most likely require manual recovery steps after a failover, except you are brave enough to try out mysql master/master (wouldn’t do)
>
>
> Am 28.02.25, 11:25 schrieb "Bryan Tiang" <bryantiang15@hotmail.com <ma...@hotmail.com> <mailto:bryantiang15@hotmail.com <ma...@hotmail.com>>>:
>
>
>
>
> Hi Community,
>
>
>
>
> We recently started to review our Disaster Recovery Strategy for Cloudstack Management, just in case something happens to the Datacenter.
>
>
>
>
> Few questions:
>
>
>
>
> #1 - Has anyone been able to get Multiple Management Servers to connect to mysql? We tried it many months ago but couldn’t get it to work.
>
>
>
>
> #2 - When Cloudstack Agents in the Hypervisor Server connects to our management, we use Private IPs. But we are wondering if it can work with subdomains instead? We are thinking of changing the Private IP, for agents to point to the DR Management Server (Synchronously replicated) when a disaster occurs. (Warm Standby DR Strategy)
>
>
>
>
> #3 - Has anyone achieved an ‘Active-Active’ Disaster Recovery Strategy? Where there are 2 Active Cloudstack Mgmt Instances in 2 Datacenter. Then, the Cloudstack Agents can load balance between the management, and auto failover to the healthy one if anyone of it goes down? Of course, we make sure CS Agent has connectivity to both Managements Private IPs.
>
>
>
>
> Regards,
> Bryan
>
>
>
>
>
>
>
>
>