Welcome to the Linux Foundation Forum!

Course content

b10s
b10s Posts: 45
edited September 2019 in LFS258 Class Forum

Seems unit file is not for kubelet

«1

Comments

  • Hi @b10s ,
    You are correct, it is for the kube-controller-manager, the previous service overseeing Kubernetes processes.

    Regards,
    -Chris

  • b10s
    b10s Posts: 45

    I'm not sure I can understand the marked phrase:

    According to the definition of bridge here https://wiki.linuxfoundation.org/networking/bridge it connects two ethernet networks. So when we create a bridge I expect rather two ethernet networks (interfaces) to be provided but ip range.

    Could you please explain more?

  • b10s
    b10s Posts: 45

    Since --export option is deprecated, what can I use to achieve the same: to get definition of resource without runtime specific uniq parameters ?

  • b10s
    b10s Posts: 45

    is there such a thing or is it a typo?

  • b10s
    b10s Posts: 45

    seems forgotten --all in the second command or name or label selector

  • b10s
    b10s Posts: 45

    Does it really call curl or is it true for specific version?

    $ sudo strace -f -e trace=network kubectl get po 2>&1
    strace: Process 29280 attached
    strace: Process 29281 attached
    strace: Process 29282 attached
    strace: Process 29283 attached
    strace: Process 29284 attached
    strace: Process 29285 attached
    strace: Process 29286 attached
    [pid 29279] socket(PF_INET, SOCK_STREAM|SOCK_CLOEXEC|SOCK_NONBLOCK, IPPROTO_IP) = 3
    [pid 29279] setsockopt(3, SOL_SOCKET, SO_BROADCAST, [1], 4) = 0
    [pid 29279] connect(3, {sa_family=AF_INET, sin_port=htons(8443), sin_addr=inet_addr("192.168.99.101")}, 16) = -1 EINPROGRESS (Operation now in progress)
    [pid 29279] getsockopt(3, SOL_SOCKET, SO_ERROR, [0], [4]) = 0
    [pid 29279] getpeername(3, {sa_family=AF_INET, sin_port=htons(8443), sin_addr=inet_addr("192.168.99.101")}, [16]) = 0
    [pid 29279] getsockname(3, {sa_family=AF_INET, sin_port=htons(41762), sin_addr=inet_addr("192.168.99.1")}, [16]) = 0
    [pid 29279] setsockopt(3, SOL_TCP, TCP_NODELAY, [1], 4) = 0
    [pid 29279] setsockopt(3, SOL_SOCKET, SO_KEEPALIVE, [1], 4) = 0
    [pid 29279] setsockopt(3, SOL_TCP, TCP_KEEPINTVL, [30], 4) = 0
    [pid 29279] setsockopt(3, SOL_TCP, TCP_KEEPIDLE, [30], 4) = 0
    strace: Process 29287 attached
    No resources found.
    [pid 29284] +++ exited with 0 +++
    [pid 29287] +++ exited with 0 +++
    [pid 29286] +++ exited with 0 +++
    [pid 29285] +++ exited with 0 +++
    [pid 29283] +++ exited with 0 +++
    [pid 29282] +++ exited with 0 +++
    [pid 29281] +++ exited with 0 +++
    [pid 29280] +++ exited with 0 +++
    +++ exited with 0 +++
    
    
  • b10s
    b10s Posts: 45
    edited September 2019


    It says GET is possible for exec.

    However my kubectl proposes me to do POST:

    Could you show an example with GET and exec ?

  • b10s
    b10s Posts: 45

  • b10s
    b10s Posts: 45

  • serewicz
    serewicz Posts: 1,000

    Thank you for your feedback.
    Regards,

  • b10s
    b10s Posts: 45

    @serewicz thanks for checking. Sorry for being noisy. Would be great to get the answers : )

    one more question:

    What does "group" mean here?

  • serewicz
    serewicz Posts: 1,000

    What have you found in your research and usage of the term?

  • b10s
    b10s Posts: 45

    Based on the context - there were mentioned only apis groups - it is obviously not what author wanted to say.
    Probably it should be autoscaling group with a short explanation of it or a link to it. Otherwise we - k8s beginners - doing LFS258 Kubernetes Fundamentals may be a bit confused

  • serewicz
    serewicz Posts: 1,000

    Thank you for your feedback.

  • b10s
    b10s Posts: 45

    I thought pod has no controller: kubelet creates it according to spec and that's all: no status check, no healing.

    What is the name of the pod's controller watch loop ?

  • b10s
    b10s Posts: 45


    since it is still configuration, .e.g namespace, I would say not an **object** configurationor whatever article to use here

  • b10s
    b10s Posts: 45


    forgotten r

  • b10s
    b10s Posts: 45


    probably you meant dev-web ?

  • b10s
    b10s Posts: 45



    however

    :'(

  • b10s
    b10s Posts: 45

  • b10s
    b10s Posts: 45

    Record would work if we run it like this:

    $ kubectl set image deploy/ghost ghost=ghost:08 --all --record=true
    deployment.extensions/ghost image updated
    
    $ kubectl rollout history deployment ghost 
    deployment.extensions/ghost 
    REVISION  CHANGE-CAUSE
    1         <none>
    2         <none>
    3         kubectl set image deploy/ghost ghost=ghost:08 --all=true --record=true
    
    
  • b10s
    b10s Posts: 45


    for some reason in my case there is no label run but app

  • b10s
    b10s Posts: 45
    edited September 2019


    will-ly place

    be-lly place

  • b10s
    b10s Posts: 45


    I was not able to see new service when I run kubectl proxy in one tab and kubectl get svc --all-namespaces in another.

    Could you explain more, how to explore this service?

  • b10s
    b10s Posts: 45
    edited September 2019


    Just curious. Assuming the "kernel" is linux kernel, what about windows server one?
    If I remember correct it was said, that windows server also may be a worker node.

  • b10s
    b10s Posts: 45

  • b10s
    b10s Posts: 45


    sorry but it is not clear why do we need to do those last steps with overriding port and redeploying ?

  • b10s
    b10s Posts: 45

  • b10s
    b10s Posts: 45

  • With all the issues posted as Comments within the same Discussion thread it is impossible to respond to every single one separately, so below I compiled a few responses for your consideration:

    Although deprecated, the --export option still works. There are a few alternatives, such as the --dry-run and --generator options.

    To expose the curl calls, besides strace, did you try running kubectl with the verbosity option described on page 5.10?

    There are installation steps mixed in with deployment configurations, which are a bit confusing, as they are unrelated. Please revise and be specific about the issue reported.

    Reports of typos and unexpected commands behavior are always welcome. However, when the exercises themselves still work with commands typed correctly, please report the specific typo, not the entire exercise.

    Some exercises purposely guide the learner to misconfigure an object (such as port 8080), then revisit a configuration step to correct the errors (port 80), so that the learner becomes familiar with possible errors, ways to identify them, and how to troubleshoot them.

    With so many exercises it may be easy to overlook information provided in an earlier exercise. Please review prior exercises carefully to find any referenced information (such as the LoadBalancer service type).

Categories

Upcoming Training