Hi guys, great video! Question regarding network configuration. Can you configure high available network with bonding, and do you have link where this can be set with ironic? We use now hybrid model (virtual masters and infra, bare metal workers). We solved bonding with machine config pool for each worker as kernel arguments. It is ok for small scale, but when workers start to grow there will be pain with upgrades. It would be great if that would be part of ironic config file. Is that possible?
Hello! At the core, they are both OpenShift so many of the capabilities will be the same. In my opinion, there are two main differences. 1) Flexibility. Deploying using a virtualization platform makes it possible to customize the size and number of nodes much easier - it's as simple as creating new VMs with the CPU/mem/etc. that's needed. 2) Raw performance. With bare metal nodes there is nothing between the application and the hardware. Despite many performance advances for virtualization, there is still some performance impact that varies based on the application. There are some other factors as well. For example, Pod density. OpenShift nodes - regardless of physical or virtual - can host a maximum of 500 Pods. If your physical nodes are very large and your Pods are relatively small, then you may not be able to fully utilize the hardware before the Pod count is too high. Additionally, the cost of entitlements may play a factor - socket-based entitlements, which are only available when deploying directly to physical servers - apply to 64 cores across 1 or 2 sockets. Hope that helps!
@@andrewsullivan1175 it would be helpful for most of us if you could do a video on installing openshift on bare metal server and on vmware with pros and cons ? Thank you
@@HemanthKumarLakshmiNarayana Thank you for the suggestion! It's been a while since we talked about install methods and platforms in depth, I think this is something we can definitely talk about in the near future.
They are, but only if you use the generic, non-integrated, deployment mode. You cannot have virtual and baremetal nodes in a "baremetal IPI integrated" environment.
Stream starts: 3:07
Poison pill operator: 13:42
Bare metal discussion: 21:32
Live demo: 37:45
Rhys mentioned at around the 38:20 time stamp that the demo was setup on a virtualized bare metal cluster. Are the bare metal VMs actually RHV VMs?
Hi guys, great video!
Question regarding network configuration. Can you configure high available network with bonding, and do you have link where this can be set with ironic? We use now hybrid model (virtual masters and infra, bare metal workers). We solved bonding with machine config pool for each worker as kernel arguments. It is ok for small scale, but when workers start to grow there will be pain with upgrades. It would be great if that would be part of ironic config file. Is that possible?
Good one
Could you please let me know use cases for using OCP 4.10 on Bare metal vs VMware ESXi 7.x? Which one is better and why?
Hello! At the core, they are both OpenShift so many of the capabilities will be the same. In my opinion, there are two main differences. 1) Flexibility. Deploying using a virtualization platform makes it possible to customize the size and number of nodes much easier - it's as simple as creating new VMs with the CPU/mem/etc. that's needed. 2) Raw performance. With bare metal nodes there is nothing between the application and the hardware. Despite many performance advances for virtualization, there is still some performance impact that varies based on the application.
There are some other factors as well. For example, Pod density. OpenShift nodes - regardless of physical or virtual - can host a maximum of 500 Pods. If your physical nodes are very large and your Pods are relatively small, then you may not be able to fully utilize the hardware before the Pod count is too high. Additionally, the cost of entitlements may play a factor - socket-based entitlements, which are only available when deploying directly to physical servers - apply to 64 cores across 1 or 2 sockets.
Hope that helps!
@@andrewsullivan1175 it would be helpful for most of us if you could do a video on installing openshift on bare metal server and on vmware with pros and cons ? Thank you
@@HemanthKumarLakshmiNarayana Thank you for the suggestion! It's been a while since we talked about install methods and platforms in depth, I think this is something we can definitely talk about in the near future.
@@andrewsullivan1175 thank you
Are a mix of bare metal worker nodes and virtual nodes supported in the same cluster?
They are, but only if you use the generic, non-integrated, deployment mode. You cannot have virtual and baremetal nodes in a "baremetal IPI integrated" environment.
@@rdoxenham1 Thanks. I thought this should be since UPI is a supported installation type.