* Fix libvirt tck test
Fix the format of config getting written in /etc/libvirt/qemu.conf
Signed-off-by: Smit Gardhariya <sgardhariya@microsoft.com>
* Add dmesg/journalctl log for libvirt community test
Add dmesg and journalctl log for libvirt community
test under after_case of the testcase: verify_libvirt_tck
Signed-off-by: Smit Gardhariya <sgardhariya@microsoft.com>
---------
Signed-off-by: Smit Gardhariya <sgardhariya@microsoft.com>
* Fix KVM Unit test result parsing
Use lisa.util methods to get the matching pattern
using named regex.
Signed-off-by: Smit Gardhariya <sgardhariya@microsoft.com>
* Fix the chmod usage for kvm unit test
Pass the named parameter instead to chmod and update_filder functions.
Order was wrong which will make chmod command to be failed.
Signed-off-by: Smit Gardhariya <sgardhariya@microsoft.com>
---------
Signed-off-by: Smit Gardhariya <sgardhariya@microsoft.com>
We have 2 points in which LISA might decide to go with a minimal shell profile:
1. When first contacting a node, issuing a 'cmd' command, checking if
'Windows' is present in the output or not (rule out Windows shell
type). A negative case will incur in either the POSIX shell type,
if 'Unknown syntax' is not thrown as output, or an early decision
for the 'minimal' shell type is taken and the shell object is set
to follow that config from there on.
2. When running the actual first lisa suite SSH command on the same
node. It turns out we've seen cases where 'Unknown syntax' is not
output at early node initialization (1), for some reason. Well, the
exception path of actual command execution will check internal node
state, that tracks if a minimal profile has been tested before or
not. If not used yet and under a command execution exception, set
the node to minimal type and try again was the logic.
It turns out the 'minification' of the node config was only done
partially, in (2). We missed the full logic of overriding its final
SSH command tokenizer.
Another missed detail is that the final tokenizer logic at the minimal
shell case was too greedy in trying to remove stray quotes. Let's keep
injecting quotes on tokens with spaces on them. That will not hurt any
legit commands from any minimal context.
* add mariner1 to unsupported list
* modify assert logic to check for any error
* refactor unspported msg
* remove duplicate platform assertion
* use Any instead of list for status_file
* skip mariner2 gen2 because it's unsupported
* add debian images to unspported images list
* verify assessment result in install api call
* use Any instead of any for unknown type
We have observed scenario where hibernation logs are not
consistently present, in those cases, we still need to verify
if Hibernation was successful. One way to do this is using uptime
* Tools: Echo: enclose file in quotes
This is to support write into a file path which has space or other
special characters.
* Tools: Modprobe: support loading module with parameters
add new parameter `parameters` to `load` and pass it to modprobe
command. This is applicable only when `modules` is not of list type.
Current Flatcar images (since 3637.0.0) contain waagent and python in
/usr/{s,}bin. Search for python in the candidate path and use that to
decide whether to call waagent directly or use the module path prefixed
with /usr/share/oem.
Signed-off-by: Jeremi Piotrowski <jpiotrowski@microsoft.com>
* skip mariner3 and add timeout to wait_operation()
* add condition to check for 1804 and 1604 for assertions
* remove unused log
* address failing pipeline, modify function to accept optional arg
* Account the new kernel config for kexec
A new kernel config, CONFIG_KEXEC_FILE is introduced in recent kernel
version. This config introduces a safer systemcall to load a kexec
kernel. Now kexec kernel can be loaded by both the old and the new
systemcalls. So, kdump test should check for either of KEXEC or
KEXEC_FILE config for determining the kernel's ability.
* Make the exception message more readable
* Update kdump.py to maintain code consistency
Currently maximize_capability breaks any test cases that rely on max_data_disk_count. Trying to manually set the value in the runbook creates a 'no quota available' error.
These changes make the default value (0,) instead of None, allowing a passed in value to merge with max_capability value.
* add OS imports
* add condition to skip sles and mariner 2
* remove unused comments and add skipexception for checking vm agent service
* remove unused code
Lili Deng
bhagyapathak
Smit Gardhariya
Gustavo Lima Chaves (from Dev Box)
Baihua Lu
paull
rydailey
feng-j678
John Feng
Chi Song (from Dev Box)
Aditya Nagesh
Purna Pavan Chandra
Jeremi Piotrowski
kanchansenlaskar
Kameron Carr
Kanchan Sen Laskar
panfengxue
Bala