Project

General

Profile

Actions

action #56132

closed

coordination #43769: [functional][y][epic] Use inter-machine dependencies to ease our review

[functional][y][timeboxed:6h] Setup inter-machine dependencies for SLE 15 for the scenarios where it makes sense

Added by riafarov over 4 years ago. Updated over 4 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
Category:
Spike/Research
Target version:
SUSE QA - Milestone 28
Start date:
2019-09-24
Due date:
2019-10-22
% Done:

100%

Estimated time:
(Total: 2.00 h)
Difficulty:

Description

Motivation

See motivation in the parent ticket. In short, if test fails even on 64-bit, for some scenarios it might be waste of time to test it in the other environments.

Acceptance criteria

  1. List of scenarios, where we can avoid unnecessary test runs is identified

Here we have description of the test suites.
https://gitlab.suse.de/riafarov/qa-sle-functional-y/blob/master/SLES_Integration_Level_Testplan.md

Suggested scope: SLE 15 only.


Subtasks 1 (0 open1 closed)

action #57311: [functional][y] Make machine inter-machine dependencies for almost identical SUTs running on same hardwareResolvedoorlov2019-09-242019-10-22

Actions
Actions #1

Updated by riafarov over 4 years ago

  • Subject changed from [functional][y] Setup inter-machine dependencies for SLE 15 for the scenarois where it makes sense to [functional][y] Setup inter-machine dependencies for SLE 15 for the scenarios where it makes sense
Actions #2

Updated by riafarov over 4 years ago

  • Subject changed from [functional][y] Setup inter-machine dependencies for SLE 15 for the scenarios where it makes sense to [functional][y][timeboxed:6h] Setup inter-machine dependencies for SLE 15 for the scenarios where it makes sense
  • Description updated (diff)
  • Status changed from New to Workable
Actions #3

Updated by JRivrain over 4 years ago

  • Assignee set to JRivrain
Actions #4

Updated by JRivrain over 4 years ago

  • Status changed from Workable to Feedback

I think the whole idea of this ticket is debatable. It it is very hard to tell where it is relevant to set inter-machines dependencies, because we tend to test things differently on different machines, and the result of a same exact test may differ anyway, for example, here we can get a soft-failure on x86 but not on aarch64 for the same test.

If the goal is to lose less time:

  • It is a bad idea to prevent tests from running in parallel.
  • In review, dependencies may force us to re-launch the failing parent and wait for its execution, even in cases where children might have worked. We would lose more time than we'd gain.

This being said, it may be interesting for limiting power consumption. But then, it would make sense to make all other architectures to depend on x86_64, and it should be done globally rather than per-test.

What makes sense IMHO is to make dependencies only for almost identical SUTs, running on the same hardware, to avoid useless concurrent runs. It leaves us with few cases:

USBinstall@uefi                          < USBinstall:64bit
btrfs_libstorage-ng@64bit-ipmi           < btrfs_libstorage-ng:64bit
btrfs_libstorage-ng@s390x-kvm-sle15      < btrfs_libstorage-ng:s390x-kvm-sle12 (currently, default for s390)
crypt_no_lvm@uefi                        < crypt_no_lvm:64bit
cryptlvm_uefi@uefi                       < cryptlvm_uefi:64bit
btrfs_libstorage-ng@s390x-zVM*           < btrfs_libstorage-ng@s390x-zVM-ctc
ext4_yast@svirt-xen*                     < ext4_yast:64bit
ext4_yast@uefi                           < ext4_yast:64bit
lvm+RAID1@*                              < lvm+RAID1:64bit
mediacheck@*                             < mediacheck:64bit
minimal+base_yast@svirt*                 < minimal+base_yast:64bit
skip_registration@s390x*                 < skip_registration@s390x-zVM-ctc
skip_registration@svirt*                 < skip_registration:64bit
skip_registration@uefi                   < skip_registration:64bit
xfs@svirt*                               < xfs:64bit
Actions #5

Updated by JERiveraMoya over 4 years ago

  • Status changed from Feedback to Resolved

Let's discuss during demo which scenarios would be interesting to configure in this way if any.

Actions #6

Updated by JERiveraMoya over 4 years ago

  • Due date set to 2019-10-08

due to changes in a related task

Actions #7

Updated by riafarov over 4 years ago

  • Due date changed from 2019-10-08 to 2019-10-22

due to changes in a related task

Actions

Also available in: Atom PDF