New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Pod should not say "pending" when a fundamental failure has occured #7968
Comments
I tend to agree. However we have the general philosophy of continually trying to get reality to move towards desired state. So for example, were you to create the missing image in your example, the system would (in theory) retry the fetch, succeed, and all would be rosy. (If the system doesn't actually do that, then that's a bug, IMO). I believe that we have the "Message" field in the status of at least some API objects to convey to the user what the last notable event regarding their thing is (in this case it should in theory tell the user something like "Failed to pull image" w.r.t. the RC or Pod status. @bgrant0607 probably has the full story behind this and should decide what the best fix, if any, is here. |
If the "event" field was visible in 'get po', that'd be a perfect solution. On Fri, May 8, 2015 at 10:27 AM, Quinton Hoole notifications@github.com
|
Use 'describe pod' instead of 'get pod' and you will be pleasantly On Fri, May 8, 2015 at 10:34 AM, David Aronchick notifications@github.com
|
I am re-assigning it to myself. @aronchick have you tried kubectl describe pod like @lavalamp suggested above? Is that meet your expectation? If not, please give us more details here, I will take a look at it. Thanks! |
Description actually did work great - i just wasn't aware of it. The On Fri, May 8, 2015 at 1:34 PM, Dawn Chen notifications@github.com wrote:
|
@aronchick, kubelet does the best effort reporting of failure reasons at a glance. The recent change in #7981 added support for surfacing image pulling failures.
Would this help you notice the failure? |
This is perfect, but my biggest issue is that I'm on a 80 character screen, and stuff gets wrapped and is tough to read. Not sure what the solution there is, but maybe intelligent wrapping? Here's how it looks on my screen - why are there two lines for one container?
|
@aronchick The first line is for pod, and the following lines with empty first two columns are for containers. We are talking about what is the best way to represent those information to the end users. Besides output layout, any other improvements we could do for you? Thanks! |
As @dchen1107 said, the first line is the pod, and the subsequent lines (without pod names) are containers. We format it this way so that the container lines can reuse the columns (e.g. STATUS, CREATED, MESSAGE). FYI, the "kubectl output is too wide" issue is #7843. |
Yes, let's discuss output format on #7843 |
I am close the issue based on above comments for now. @aronchick please re-open the issue if we didn't answer your question here. Thanks! |
Repro:
Expected:
Actual:
Generally, we should have a different message/state when it's a known bad issue. Here's the Kubelet.log:
The text was updated successfully, but these errors were encountered: