kubelet: cgroups: be verbose about validation #108568
Conversation
k8s-ci-robot
added
the
release-note-none
k8s-ci-robot
added
size/M
|
/sig node |
k8s-ci-robot
added
sig/node
|
@rphillips @mrunalp this lgtm but wanted someone more familiar with cgroup manager (if that's either of you two, or others) |
| if !cgroupManager.Exists(cgroupRoot) { | ||
| return nil, fmt.Errorf("invalid configuration: cgroup-root %q doesn't exist", cgroupRoot) | ||
| if err := cgroupManager.Validate(cgroupRoot); err != nil { | ||
| return nil, fmt.Errorf("invalid configuration: %w", err) |
There was a problem hiding this comment.
NB: this is terminal to the kubelet process and not explaining what error caused the configuration to be invalid is poor UX
stevekuznetsov
force-pushed
the
skuznets/verbose-error
branch
2 times, most recently
from
7cb3880
to
0e475f3
Compare
|
/assign @kolyshkin |
|
/triage accepted |
k8s-ci-robot
added
triage/accepted
There was a problem hiding this comment.
Is it possible to not change the API (I don't like two methods with different names that do the same thing, with the only difference is return value)? Say, increase the log level in in Exists to a warning, and add logging to the missing places?
If not, maybe it's better to change all instances of if cm.Exists() to if cm.Validate() == nil.
|
Also, ideally I'd like this to go on top of #107149 since it incorporates some non-trivial changes. |
|
@kolyshkin no. I very strongly disagree with you. Logging critical errors that cause the process to exit at a high verbosity level and asking the user to re-run their setup in order to see the error should not be how we go about this. Also, if you notice, there were places where the logging did not expose the reason for |
|
@kolyshkin also while I understand the request re: the other PR, it looks very large, very old and it's not clear the timeline on which it would merge. |
stevekuznetsov
force-pushed
the
skuznets/verbose-error
branch
2 times, most recently
from
d02257c
to
1296562
Compare
k8s-ci-robot
added
area/test
sig/testing
|
@rphillips @kolyshkin updated to entirely remove the |
|
This is changing and removing an established API. Exists() should probably stay and Validate be added. Exists() can simply call Validate(). The internal code can call Validate() to propagate the errors. |
|
Gotcha. @rphillips that was the original factoring. Let me revert to that one... |
Previously, callers of `Exists()` would not know why the cGroup was or was not existing. In one call-site in particular, the `kubelet` would entirely fail to start if the cGroup validation did not succeed. In these cases we MUST explain what went wrong and pass that information clearly to the caller. Previously, some but not all of the reasons for invalidation were logged at a low log-level instead. This led to poor UX. The original method was retained on the interface so as to make this diff small. Signed-off-by: Steve Kuznetsov <skuznets@redhat.com>
|
@rphillips reverted to the original state |
stevekuznetsov
force-pushed
the
skuznets/verbose-error
branch
from
1296562
to
8f2bc39
Compare
|
/lgtm |
k8s-ci-robot
added
the
lgtm
|
/assign @derekwaynecarr |
|
/approve |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: smarterclayton, stevekuznetsov The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
k8s-ci-robot
added
the
approved
Previously, callers of
Exists()would not know why the cGroup was orwas not existing. In one call-site in particular, the
kubeletwouldentirely fail to start if the cGroup validation did not succeed. In
these cases we MUST explain what went wrong and pass that information
clearly to the caller. Previously, some but not all of the reasons for
invalidation were logged at a low log-level instead. This led to poor
UX.
Signed-off-by: Steve Kuznetsov skuznets@redhat.com
/kind bug
/cc @sjenning @derekwaynecarr @smarterclayton