Hi, I've got some quick suggestions based on my personal experience written bellow each question. Hope you find it useful. Regards, Gene Kuo On Thursday, February 22nd, 2024 at 10:19 PM, KK CHN <kkchn.in@gmail.com> wrote:
List,
I am going ahead to adopt OpensStack for our Live deployment sites which all are running previously on proprietary HCI solutions.
Few queries.
1. Which base Operating System need to be selected for production? ( prefer OpenSource community Operating Systems.) Is Debian a good choice? If so, which version ? Or any other suggestions?
It kinds of depends on the deployment tools of your choice. e.g. Kolla-Ansible has a supported operating system list. https://docs.openstack.org/kolla-ansible/latest/user/support-matrix.html
2. What backend / SDS needs to be configured for a crash resistant FaultTolerant cluster. In case of a node / Disks failure which can withstand the workload and most preferably avoid data loss in the worst case scenario.
3. Heard about Ceph. How its a best choice for a 24x7 mission critical cluster? kindly advise.
Ceph is probably the most common one according to the user survey result.
If so, what configuration architecture needs to be considered ?
3. My luxury is only at an On Prem Server Cluster comprising a minimum 3 to max 4 Nodes.
( possible DELL 650 or 750 series OR HPE servers ) .
4. As my cluster resources are limited to 3 to 4 server nodes / site ( 2 X32C 64 Threads, Intel Xeon Gold processors, with min 512 GB Ram , 4 x3.94TB SSD or 8 x1.92TB in each node)
Is it reliable/advisable to set Controllers, Compute and Storage all together in these three or four nodes?
Which is the best possible deployment options I can have with these 3/4 nodes. Kindly shed some light .
I've been doing so in my own cluster for quite a long time and didn't find any big issue for a couple years though the usage isn't high. Just note that controller nodes should be in odd number.
5. Any other points need to be considered while this must be a reliable infrastructure for running a mission critical 24x7 operation.
OpenStack services should be quite reliable running in a HA setup, but your VMs is still running on each host. If any of the host is down it will also bring down the VMs on that host. There are projects that Masakari that provides Virtual Machine High Availability but I'm not very familiar with it as I prefer implementing HA on the application on top of the VMs
6. Which mode of installation I can choose, the Kolla-ansible multinode is advisable for a production deployment ? ( I am familiar with the kolla-ansible all-in-one installation ) OR ther a better deployment method for production sites ? )
Kolla-Ansible should work, there are existing cases that companies use it for production deployment. I would say choose the deployment method that you are most familiar with for easier troubleshooting.
Kindly shed some light and your valuable suggestions . Any hints most welcome.
Thank you, Krish