docs: Manually create Equinix cluster #2
Conversation
Signed-off-by: Ross Fairbanks <[email protected]>
| export PROJECT_ID="<PROJECT_ID>" | ||
|
|
||
| # Use Paris metro (Equinix region) | ||
| export METRO="pa" |
There was a problem hiding this comment.
Do we want to consider carbon intensity when deciding which metro to use?
I chose Paris since France has a low average carbon intensity and this metro has a good selection of node types.
Full list is here and we can easily change if I'm over thinking this :)
https://deploy.equinix.com/developers/docs/metal/locations/metros/
There was a problem hiding this comment.
This is a very important callout! We can hardcode Paris for now and make the testing carbon-aware in the future. Would you like to create a new issue for this to add it to our backlog? 🌟
There was a problem hiding this comment.
Yes sure I'll create an issue for it. 👍
|
|
||
| # Use node type with Intel CPUs for RAPL support | ||
| export CONTROLPLANE_NODE_TYPE="m3.small.x86" | ||
| export WORKER_NODE_TYPE="m3.small.x86" |
There was a problem hiding this comment.
I think its safest to use a node type with an Intel CPU. I tested with an AMD CPU and Kepler couldn't access RAPL.
I think this node type has good enough resources for now. WDYT?
There was a problem hiding this comment.
+1 for using an Intel CPU and starting with small! 👍
| apt install linux-headers-$(uname -r) | ||
| apt install linux-modules-extra-$(uname -r) | ||
| modprobe intel_rapl_common | ||
| ``` |
There was a problem hiding this comment.
I didn't actually need to run these steps but I think its safest to include them since they are requirements for Kepler. WDYT?
There was a problem hiding this comment.
Good to validate that the kernel headers are there, I always double-check this as well 👍
Signed-off-by: Ross Fairbanks <[email protected]>
rossf7
force-pushed
the
add-equinix-metal-docs
branch
from
efe696d to
b915481
Compare
Signed-off-by: Ross Fairbanks <[email protected]>
Adds docs for manually creating a cluster with CAPI / CAPEM to try things out while we work on the pipeline design.
I've added comments for decisions I'm not sure on or that are worth highlighting.
One thing to note is the Kubernetes API will be public. As we've discussed as a later step we could make it private and use a GitOps tool like Flux to manage the components running in the cluster.