Bug 1774942 - Booting from image causes OCI runtime error, "context deadline exceeded"
Summary: Booting from image causes OCI runtime error, "context deadline exceeded"
Keywords:
Status: CLOSED WONTFIX
Alias: None
Product: Fedora
Classification: Fedora
Component: kata-runtime
Version: 32
Hardware: Unspecified
OS: Unspecified
unspecified
unspecified
Target Milestone: ---
Assignee: Christophe de Dinechin
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2019-11-21 10:21 UTC by Christophe de Dinechin
Modified: 2020-04-21 10:23 UTC (History)
3 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed: 2020-04-21 10:23:36 UTC
Type: Bug
Embargoed:


Attachments (Terms of Use)

Description Christophe de Dinechin 2019-11-21 10:21:02 UTC
Description of problem:

Booting with 'image' configuratoin instead of 'initrd' results in OCI runtime error:

% podman run --security-opt label=disable --runtime /usr/bin/kata-runtime -it alpine sh
Error: Failed to check if grpc server is working: context deadline exceeded: OCI runtime error


Version-Release number of selected component (if applicable): kata-runtime-1.8.2-4.fc31.x86_64, kata-runtime-1.9.0-1.fc31.x86_64


How reproducible: Always


Steps to Reproduce:
1. Edit /usr/share/kata-containers/defaults/configuration.toml to replace

  initrd = "/usr/share/kata-containers/kata-containers-initrd.img"
  #image = "/usr/share/kata-containers/kata-containers.img"

with

  #initrd = "/usr/share/kata-containers/kata-containers-initrd.img"
  image = "/usr/share/kata-containers/kata-containers.img"

2. Run podman, e.g.

  podman run --security-opt label=disable --runtime /usr/bin/kata-runtime -it alpine sh


Actual results:

Error: Failed to check if grpc server is working: context deadline exceeded: OCI runtime error


Expected results:

Image boots as with initrd

Additional info:

Comment 1 Christophe de Dinechin 2019-11-21 10:21:58 UTC
Hey Cole, I wonder if you had tested this configuration

Comment 2 Christophe de Dinechin 2019-11-21 10:30:25 UTC
Ouptut of dmesg with `enable_debug=true` is not that informative:

[2818435.190330] IPv6: ADDRCONF(NETDEV_CHANGE): vethf3835a1d: link becomes ready
[2818435.190409] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready
[2818435.196248] cni-podman0: port 1(vethf3835a1d) entered blocking state
[2818435.196250] cni-podman0: port 1(vethf3835a1d) entered disabled state
[2818435.196322] device vethf3835a1d entered promiscuous mode
[2818435.196396] cni-podman0: port 1(vethf3835a1d) entered blocking state
[2818435.196397] cni-podman0: port 1(vethf3835a1d) entered forwarding state
[2818435.438314] eth0: Caught tx_queue_len zero misconfig
[2818438.450773] cni-podman0: port 1(vethf3835a1d) entered disabled state
[2818438.456599] device vethf3835a1d left promiscuous mode
[2818438.456608] cni-podman0: port 1(vethf3835a1d) entered disabled state

Comment 3 Cole Robinson 2019-11-21 13:05:40 UTC
Yes I tested this config, as mentioned before this is what the kernel nvdimm request is about: 

https://bugzilla.redhat.com/show_bug.cgi?id=1750581
https://bugzilla.redhat.com/show_bug.cgi?id=1696481

Without those modules built into the kernel, image= has no chance of working at the moment.
That's likely the root issue.

Unfortunately debugging these things with kata is a real pain.
The way I did it, was to extract a qemu command to run manually, then I had easier options
to capture appliance boot output, pass kernel arguments, also pass a custom initrd with
dracut-systemd and systemd-initrd modules, and the 'rescue' boot option, which gives
a way to inspect the appliance state.

Comment 4 Ben Cotton 2020-02-11 17:45:17 UTC
This bug appears to have been reported against 'rawhide' during the Fedora 32 development cycle.
Changing version to 32.

Comment 5 Fabiano Fidêncio 2020-04-21 10:23:36 UTC
As on Fedora we've decided to not use the image, but only the initrd method (see https://src.fedoraproject.org/rpms/kata-osbuilder/c/bd4294598d9dfb5807b8b2490c870dd2ebf9cf32?branch=master) we can close this one as WONTFIX.

If there's the need to revisit this in the future, let's just re-open it without any issue.


Note You need to log in before you can comment on or make changes to this bug.