action #111710
closed[qa-tools] [tools] remove usage of *_TEST_TEMPLATE vars in qem-bot and openQA media definition if favor of *_TEST_REPOS size:M
0%
Description
Current state:¶
Now QEM uses in aggregate jobs pretty complicated system to add repositories under test to jobs itself:
qem-bot post jobs with *_TEST_ISUES variable.
In osd are defined media with templates in corresponding *_TEST_TEMPLATE
when is job started from this vars is calculated variable MAINT_TEST_REPO using SCC_ADDONS
used code for this transformation:
main_common.pm -> https://github.com/os-autoinst/os-autoinst-distri-opensuse/blob/master/lib/main_common.pm#L773-L789
SLE main.pm -> https://github.com/os-autoinst/os-autoinst-distri-opensuse/blob/master/products/sle/main.pm#L323-L373
Cons:¶
- pretty complex settings for medium with high probability of error
- code used for this in os-autoinst-distri-opensuse is pretty complex
Proposal:¶
modify qem-bot to post new vars (*_TEST_REPO) with corresponding repositories and the use simpler code to construct MAINT_TEST_REPO by simply joining this vars still using SCC_ADDONS to join correct products/modules
qem-bot has all needed info for generate this vars which removes one layer of possible human errors (still can be problem in qa-metadata, but it is simpler to spot and debug)
Pro's:¶
- remove mess from media settings
- cleaner/simpler code in os-autoinst-distri-opensuse
Cons:¶
- lights more complex code in qem-bot ( Aggregate class )
- longer command line presented by qem-bot for aggregates
Acceptance criteria¶
AC1: implement needed changes in qem-bot
AC2: implement needed changes in os-autoinst-distri-opensuse
AC3: (optional) cleanup media definitions in OSD
O3¶
Note: this changes only handling for OSD, O3 uses different bot so poo is not related ( but can be also implemented ? on O3 side)
Updated by jbaier_cz over 1 year ago
That sounds like a good idea. I believe the bot can handle it without adding much complexity. There is also yet another bonus, without the calculation in "runtime" it should be easier and more transparent to clone a job with different subset if needed.
Updated by okurz over 1 year ago
- Target version set to future
I agree with the problem statement. Not sure if moving more complexity into qem-bot is good though.
Updated by osukup over 1 year ago
okurz wrote:
I agree with the problem statement. Not sure if moving more complexity into qem-bot is good though.
in reality for qem-bot increase of complexity is pretty minimal --> needs only modify this https://github.com/openSUSE/qem-bot/blob/master/openqabot/types/aggregate.py#L167-L168 loop ( +- 2-3 lines of code ?)
on other side it removes from os-autoinst-distri more than 20 lines of code and many more from OSD media definitions
Updated by osukup over 1 year ago
jbaier_cz wrote:
That sounds like a good idea. I believe the bot can handle it without adding much complexity. There is also yet another bonus, without the calculation in "runtime" it should be easier and more transparent to clone a job with different subset if needed.
there still be some calculation online, but less complex
Updated by osukup over 1 year ago
- Subject changed from [qa-tools] remove usage of *_TEST_TEMPLATE vars in qem-bot and openQA media definition to [qa-tools] remove usage of *_TEST_TEMPLATE vars in qem-bot and openQA media definition if favor of *_TEST_REPOS
Updated by dvenkatachala over 1 year ago
- Related to action #109707: [qe-core] Detect when a maintenance update has been already released before running the test added
Updated by osukup over 1 year ago
@okurz @cdywan we should increase priority / add it in workable ( as handling this vars is pretty big part of poo#109707 ) and we don't want duplicate complex code in os-autoinst-distri-openSUSE + openQA itself.
Updated by osukup over 1 year ago
- Subject changed from [qa-tools] remove usage of *_TEST_TEMPLATE vars in qem-bot and openQA media definition if favor of *_TEST_REPOS to [qa-tools] [tools] remove usage of *_TEST_TEMPLATE vars in qem-bot and openQA media definition if favor of *_TEST_REPOS
Updated by okurz over 1 year ago
Sure, as soon as we have the current "normal" in the backlog resolved this would be one candidate
Updated by livdywan over 1 year ago
- Status changed from New to Feedback
- Assignee set to osukup
- Target version changed from future to Ready
Updated by osukup over 1 year ago
- Description updated (diff)
- Status changed from Feedback to In Progress
qem-bot part implemented.
for full use it needs changes in os-autoinst-distri-openSUSE
Updated by openqa_review over 1 year ago
- Due date set to 2022-08-23
Setting due date based on mean cycle time of SUSE QE Tools
Updated by okurz over 1 year ago
https://openqa.suse.de/tests/9296092#settings is a job using the new format with the _TEST_REPOS variables. Could you please check that the according code of https://github.com/os-autoinst/scripts/pull/177 still works with the new format? Maybe it's enough to simply extend https://github.com/os-autoinst/scripts/blob/master/openqa-trigger-bisect-jobs#L76 to look for _TEST_(ISSUES|REPOS) and maybe the sorting and grouping already works automatically even with repositories
Updated by mkittler over 1 year ago
- Subject changed from [qa-tools] [tools] remove usage of *_TEST_TEMPLATE vars in qem-bot and openQA media definition if favor of *_TEST_REPOS to [qa-tools] [tools] remove usage of *_TEST_TEMPLATE vars in qem-bot and openQA media definition if favor of *_TEST_REPOS size:M
Updated by osukup over 1 year ago
- Status changed from In Progress to Feedback
merged changes in os-auinst-distri-opensuse
so now AC1 and AC2 fullfilled
Updated by osukup over 1 year ago
okurz wrote:
Please look into #111710#note-13
bisect code needs some adjustment , on OSD are now TEST_ISSUES irrelevant + in reality code itself now don't work as excepted. Because it not undefine MAINT_TEST_REPO variable. If its defined (and in clone by default is defined ) whole *_TEST magic is skipped
Updated by osukup over 1 year ago
https://github.com/os-autoinst/scripts/pull/182 -- add support for REPOS and fixed issue with MAINT_TEST_REPOS
unfortunately it has still faulty logic while triggering jobs...
Updated by ph03nix over 1 year ago
This change broke our PublicCloud ARM64 test runs by changing the architecture of the _TEST_REPO to the architecture of the helper VM instead of using the *_TEST_TEMPLATE.
See e.g. https://openqa.suse.de/tests/9369145#settings where BASE_TEST_REPOS is x86_64, while it should be aarch64 (as previously defines in the BASE_TEST_TEMPLATE).
https://github.com/openSUSE/qem-bot/pull/65 introduces a temporary workaround while we search for a better solution. The issue is handled in #115652
Addendum: metadata change: https://gitlab.suse.de/qa-maintenance/metadata/-/merge_requests/681
Updated by osukup over 1 year ago
ph03nix wrote:
This change broke our PublicCloud ARM64 test runs by changing the architecture of the
_TEST_REPOto the architecture of the helper VM instead of using the*_TEST_TEMPLATE.
See e.g. https://openqa.suse.de/tests/9369145#settings whereBASE_TEST_REPOSisx86_64, while it should beaarch64(as previously defines in theBASE_TEST_TEMPLATE).
prety nonstandard use of _TEST_TEMPLATE..
https://github.com/openSUSE/qem-bot/pull/65 introduces a temporary workaround while we search for a better solution. The issue is handled in #115652
ook, workaround meged , have fun
Updated by livdywan over 1 year ago
- Due date changed from 2022-08-23 to 2022-09-02
Considering the impact on e.g. PC tests and limited team capacity I'd say we give this some more time and see where we're at early next week.
Updated by osukup over 1 year ago
- Status changed from Feedback to Resolved
It works more than two weeks without known problems (except QAC case which was caused by 'non-standard' and practically incorrect usage of templates)*
- workaround from QAC merged into qem-bot