Failed to create the kube client with proxy

[vitality411]

Hello,
when I use kubectl-view-allocations in an environment with http_proxy and https_proxy variables I get the following error:

k view-allocations -g node
Error:
   0: Failed to create the kube client
   1: configured proxy http://squid:3128/ requires the disabled feature "kube/http-proxy"

Location:
   src/main.rs:47

Backtrace omitted. Run with RUST_BACKTRACE=1 environment variable to display it.
Run with RUST_BACKTRACE=full to include source snippets.

When I unset the variables, it works without issues. Is there some way to circumvent this?

Thanks

[davidB]

Thank you for reporting (with the log).
I'm releasing version 0.20.3 with http-proxy enabled.
Please re-open if you still have the issue.

[vitality411]

@davidB I don't have permissions to reopen.

I now get the following error:

RUST_BACKTRACE=1 RUST_BACKTRACE=full kubectl view-allocations -g node -u
2025-01-17T05:45:38.104193Z ERROR failed with error client error (Connect)
Error:
   0: Failed to list nodes
   1: ServiceError: client error (Connect)
   2: client error (Connect)
   3: deadline has elapsed

Location:
   src/main.rs:47

  ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ BACKTRACE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
  <empty backtrace>

Run with COLORBT_SHOW_HIDDEN=1 environment variable to disable frame filtering.

[davidB]

It's a permission issue, do you have the permissions to list nodes with the current kube's config.

kubectl get nodes

[vitality411]

Yes, I have permissions to list nodes. After I do

unset HTTPS_PROXY
unset https_proxy

it works fine but not as long these envs are set.

[upszot]

Hello, I'm running the latest version available on Krew as of today, which is 0.21.1.
I am also in an environment with a Proxy configured.

The error it returns is this:

`
2025-05-26T14:25:20.635491Z ERROR failed with error client error (Connect)
Error:
0: Failed to list nodes
1: ServiceError: client error (Connect)
2: client error (Connect)
3: deadline has elapsed

Location:
src/main.rs:41
`

And just like @vitality411, it works correctly after running:

unset HTTPS_PROXY
unset https_proxy

[davidB]

Maybe a stupid question,
From my research, kubectl seems to ignore HTTPS_PROXY in a lot of cases (eg, loopback), and the documentation only mentions proxy for socks5. I also saw examples where HTTPS_PROXY is set, but IP addresses of the k8s cluster are listed in the NO_PROXY environment variable.
Are you sure that the proxy allows access to the Kubernetes cluster?

I ask because you seem to be able to access the cluster without a proxy. SO maybe the issue is that the plugin honors/uses HTTPS_PROXY.

[upszot]

I'll give you more details.

• Yes, the proxy allows access to the cluster.
• I have the cluster IPs loaded in the "no_proxy" environment variable.

[davidB]

Thank you @upszot, but can you try 'unset HTTP_PROXY HTTPS_PROXY ...' then run k topnodes?

To confirm that kubectl use it or not (it's not clear from the documentation, and golang reimplement part of the network layer).

If without HTTP_PROXY,NO_PROXY,... kubectl ... continues to work, it means that it doesn't care of those variables and completely ignores them (vs the plugin).

[upszot]

Hello, sorry for the delay. I'm sending you the information you requested:

• When the proxy variables are set:

       - The command kubectl top nodes works without a problem.
       - The command kubectl view_allocations throws the described error.
  

• When unsetting HTTP_PROXY, http_proxy, HTTPS_PROXY, and https_proxy:

      - The command kubectl top nodes continues to work without problems.
      - The command kubectl view_allocations starts working.
  

I've attached the outputs for greater clarity.

Regards,

`
[user@server ~]$ env |grep -i proxy
HTTP_PROXY=/
FTP_PROXY=/
https_proxy=/
http_proxy=/
no_proxy=*****
NO_PROXY=*****
HTTPS_PROXY=/
ftp_proxy=/
[user@server ~]$ k top nodes
NAME CPU(cores) CPU(%) MEMORY(bytes) MEMORY(%)
node-010 1359m 33% 2932Mi 63%
node-012 254m 6% 3223Mi 69%
node-013 470m 11% 3040Mi 65%
node-004 3897m 24% 13848Mi 43%
node-005 4351m 27% 14583Mi 45%
node-006 2617m 16% 13239Mi 41%
node-007 78m 0% 1668Mi 5%
node-008 760m 4% 2284Mi 7%
[user@server ~]$ unset HTTP_PROXY https_proxy HTTPS_PROXY https_proxy
[user@server ~]$ k top nodes
NAME CPU(cores) CPU(%) MEMORY(bytes) MEMORY(%)
node-010 1396m 34% 2930Mi 63%
node-012 311m 7% 3205Mi 69%
node-013 451m 11% 3045Mi 66%
node-004 3268m 20% 13836Mi 43%
node-005 4904m 30% 14557Mi 45%
node-006 1796m 11% 13209Mi 41%
node-007 76m 0% 1669Mi 5%
node-008 254m 1% 2241Mi 7%
[user@server ~]$ k view_allocations
Resource Requested Limit Allocatable Free
cpu (9%) 8.5 (5%) 5.0 92.0 83.5
├─ node-010 (25%) 1.0 __ 4.0 3.0
│ ├─ calico-node-tvthz 250.0m __ __ __
│ ├─ coredns-84994b8c4-z8clr 100.0m __ __ __
│ ├─ etcd-node-010 100.0m __ __ __
│ ├─ kube-apiserver-node-010 250.0m __ __ __
│ ├─ kube-controller-manager-node-010 200.0m __ __ __
│ └─ kube-scheduler-node-010 100.0m __ __ __
├─ node-012 (25%) 1.0 __ 4.0 3.0

`
I just tested this on version kubectl-view-allocations 1.0.0 and the problem still persists.

[davidB]

Thank you for the feedback.

It confirms what I read in some places: kubectl ignores HTTP_PROXY, http_proxy, HTTPS_PROXY, and https_proxy.

I'll look at the underlying library used to manage the connection to k8s, if I can provide an option to ignore the http_proxy environment variables (on demand).