-
Notifications
You must be signed in to change notification settings - Fork 63
Remove all the patches removing ulimit calls
#1040
Conversation
mook-as
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'd like to see the commit messages to actually include reasons the patches are safe to remove (and why we're removing them), rather than just saying that they are, next time.
bosh/releases/pre_render_scripts/api/cloud_controller_ng/jobs/patch_cc_uploader_ctl.sh
Show resolved
Hide resolved
bosh/releases/pre_render_scripts/api/cloud_controller_ng/jobs/patch_cloud_controller_api_ctl.sh
Show resolved
Hide resolved
bosh/releases/pre_render_scripts/api/cloud_controller_ng/jobs/patch_file_server_ctl.sh
Show resolved
Hide resolved
bosh/releases/pre_render_scripts/auctioneer/auctioneer/jobs/patch_auctioneer_ctl.sh
Show resolved
Hide resolved
bosh/releases/pre_render_scripts/diego-api/locket/jobs/patch_locket_ctl.sh
Show resolved
Hide resolved
bosh/releases/pre_render_scripts/router/gorouter/jobs/patch_pre-start.sh
Show resolved
Hide resolved
bosh/releases/pre_render_scripts/scheduler/ssh_proxy/jobs/patch_ssh_proxy_ctl.sh
Show resolved
Hide resolved
|
Thank you for the reminder about the commit message. Will fix. |
… calls from a pre-start and various other scripts. With the exception of (`router/gorouter/jobs/patch_pre-start.sh`) none of the scripts they change are used. They are old-style start/stop control scripts ignored by the operator (which uses the BPM information instead). The pre-start script TOH is used, however as the change happens in a different container (versus the job runtime) it is not doing anything useful, either way.
43fd7f2 to
5f7cd28
Compare
|
Squashed history and amended the commit message based on Mark's comments. |
mook-as
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I believe the problem wasn't that the ulimit calls weren't doing anything useful, it's that they failed (because we don't have the permissions to make things higher), causing the whole thing to fall over.
It appears that the operator gives us containers with big enough preset limits that we no longer fail the calls.
Description
Removed all the patches which removed
ulimitcalls from various pre-start and other kubecf scripts.Motivation and Context
This is in the context of tickets #504 and #505, where a first set of these patches was given to upstream for consideration. In the recent check of what they were used for the discussion introduced the possibility that the patches might not be necessary anymore, or even if, that the underlying issue could be solved differently, via ops files modifying job capabilities.
Testing without the patches strongly indicated them as not necessary anymore. Thus this PR removes them, and the relevant upstream PRs (cloudfoundry/capi-release#172, cloudfoundry/diego-release#512) will be retracted and closed.
How Has This Been Tested?
It was tested that both then-master f90161a and the branch from it without the ulimit patches deploy and then pass the 🐈 tests.
The initial testing using an SA config on a local minikube with 16/20 G memory failed severely in the 🐈 even for the unchanged baseline, making this attempt useless.
The final testing then happened using a GKE cluster deployed via catapult. The necessary bundle of charts was created from the relevant commit via
The catapult configuration used was
using the targets
Types of changes
Checklist: