feat: install nvidia driver using runfile #482
Conversation
|
Minimum allowed coverage is Generated by 🐒 cobertura-action against cefc034 |
faiq
force-pushed
the
faiq/nvidia-runfile
branch
2 times, most recently
from
e4b3f1f
to
92adde0
Compare
faiq
force-pushed
the
faiq/nvidia-runfile
branch
from
fc0b460
to
2160e81
Compare
| $(MAKE) devkit.run WHAT="make build-$* \ | ||
| BUILD_DRY_RUN=${BUILD_DRY_RUN} \ | ||
| VERBOSITY=$(VERBOSITY) \ | ||
| ADDITIONAL_ARGS="--instance-type=g4dn.2xlarge$(if $(ADDITIONAL_ARGS),$(SPACE)$(ADDITIONAL_ARGS))" \ |
There was a problem hiding this comment.
this instance is what kaptain uses for their testing. we should do also test on this regularly. we've seen issues with this in the past
|
|
||
| # to pick up the changes to unload nouveau | ||
| # some hardware like aws g4dn.2xlarge require this | ||
| - name: unconditionally reboot the machine with all defaults |
There was a problem hiding this comment.
There was a problem hiding this comment.
Will the machine also reboot for preprovisioned, or is sysprep==false in that case?
There was a problem hiding this comment.
it's false.
however, they might need to reboot it on their own depending on OS if they were previously using nouveu drivers
| - name: run nvidia-smi | ||
| shell: | ||
| executable: /bin/bash | ||
| cmd: nvidia-smi |
There was a problem hiding this comment.
it should add it to the path on install.
|
|
||
| # to pick up the changes to unload nouveau | ||
| # some hardware like aws g4dn.2xlarge require this | ||
| - name: unconditionally reboot the machine with all defaults |
There was a problem hiding this comment.
Will the machine also reboot for preprovisioned, or is sysprep==false in that case?
| apt: | ||
| name: linux-headers-{{ ansible_kernel }} | ||
| name: | ||
| - linux-headers-{{ ansible_kernel }} |
There was a problem hiding this comment.
$ apt-cache depends build-essential
build-essential
|Depends: libc6-dev
Depends: <libc-dev>
libc6-dev
Depends: gcc
Depends: g++
Depends: make
make-guile
Depends: dpkg-dev
There was a problem hiding this comment.
makes it a little cleaner :)
faiq
changed the title
faiq
changed the title
faiq
changed the title
|
just want to make clear that this patch deleted the vault functionality to make sure staying with the centos minor version kernel. That means whenever RedHat releases a 7.10 release we'll either make a half 7.9 half 7.10 image as we'll install 7.10 patch in a 7.9 image or we'll simply have to immediately drop support for 7.9 to follow 7.10. But the minor version is hard coded into every KIB build. So @faiq I really really suggest keeping this backward compatibility as we render previous KIB versions unusable as soon as there is a new RHEL7/CentOS7 release CentOS moves almost instant the kernel packages of 7.9 into vault once a new release is there. This is the reason for the vault code as it drops the dependency on CentOS releases and timeline |
What problem does this PR solve?:
Starting work on installing drivers with runfiles
Which issue(s) does this PR fix?:
Special notes for your reviewer:
Does this PR introduce a user-facing change?: