action #11432
closedFeature 312751: snapper: cleanup rules based on free space and/or fill-level
Added by maritawerner over 8 years ago. Updated almost 8 years ago.
100%
Description
For details see https://fate.suse.com/312751
First check if the Feature status is "done".
Description:
During a discussion in the Beta tester group we found that it would be useful not only cleanup btrfs snapshots based on "age", but also depending on the filesystem's fill-level and/or free space
Files
Updated by dgutu over 8 years ago
Testcase for this feature:
http://crocodile.qa.suse.cz/tests/3105
Updated by dgutu over 8 years ago
- % Done changed from 0 to 30
Test is present but I'd like to have a review from @Rich
Updated by dgutu over 8 years ago
- % Done changed from 30 to 80
PR is done but Olive suggested to try the test on TW.
TW by default sets separate /home and root so for my test the / of 10 Gb is too small.
On SLES /home resides on / and test is running fine.
Suggestions?
Updated by okurz over 8 years ago
PR in question: https://github.com/os-autoinst/os-autoinst-distri-opensuse/pull/1386
only enable the tests for SLES then for now.
Updated by okurz over 8 years ago
- Target version changed from 168 to Milestone 3
Updated by dgutu over 8 years ago
Something doesn't work in this snapper cleanup algorithm.
Could be that the HDD of 20Gb is too small. I will try with 30 and update the task.
Updated by dgutu over 8 years ago
dgutu wrote:
Something doesn't work in this snapper cleanup algorithm.
Could be that the HDD of 20Gb is too small. I will try with 30 and update the task.
http://crocodile.qa.suse.cz/tests/3775#step/snapper_cleanup/71
Updated by dgutu over 8 years ago
- File download.png download.png added
I've found another scenario when this test shouldn't be run.
The HDD images we have using for migration already are full of important snapshots
so deleting them will broke another test to run.
Updated by okurz over 8 years ago
dgutu I think you might be wanting too much in one step :-) Better just add the test for one single scenario than not have it ready at all.
Feature test: PASSED
Used dgutu's PR, adapted it a bit and ran in
http://lord.arch/tests/2682
but it takes rather long to execute so I am convinced it must not be called as part of every console test.
I tried to execute it as part of "extra tests" but it turns the filesystem R/O, probably because space got depleted? See http://lord.arch/tests/2687#step/snapper_cleanup/29
Even worse, it doesn't fail the test.
But nevertheless, here are my suggestions: https://github.com/dgutu-suse/os-autoinst-distri-opensuse/compare/features/snapper_cleanup...okurz:feature/snapper_cleanup?expand=1
Updated by dgutu about 8 years ago
Bug was created - https://bugzilla.suse.com/show_bug.cgi?id=998360
Updated by dgutu about 8 years ago
Test was added to o.s.d, for now it will part of sles+extratests.
Updated by dgutu about 8 years ago
I need your opinion on the following:
- The bug created based on behavior of snapper_cleanup test could be invalid IF the root partition where test is running is not less then 20GB.
- Every run where I have used HDDSIZEGB=22 and TOGGLEHOME=1 passed.
Should I consider an update on testcase 'sles+extratests' by using HDDSIZEGB=22 and
TOGGLEHOME=1 (the default layout will create a separate /home and / if the last will be used)?
Thx.
Updated by maritawerner about 8 years ago
Dumitru, could you please check with Martin Kravec about qgroups? Maybe he can check here.
Updated by dgutu about 8 years ago
maritawerner wrote:
Dumitru, could you please check with Martin Kravec about qgroups? Maybe he can check here.
We already had a discussion with Martin, our tasks are different.
for x86_64 snapper cleanup is passing but it behaves different for ppc for unknown reason for now, I will investigate.
Updated by dgutu about 8 years ago
- Status changed from Feedback to Resolved
This feature is ready and works in good way in x86_64.
There is still some issue on ppc64le platform but this I will rise a bug.
Working example: https://openqa.suse.de/tests/640886#
I will mark this as resolved.
Updated by dgutu about 8 years ago
- Status changed from Resolved to In Progress
- % Done changed from 100 to 90
Updated by okurz almost 8 years ago
I guess you set it to "in progress" because of the differing behaviour on ppc64le, right? can you shortly mention the corresponding bug?
Updated by dgutu almost 8 years ago
Because on ppc architecture the block size is much bigger than on x86 architecture I've decided to create separate testcase for ppc.
So, I've added 2 testcase, one for creating the hdd image with bigger size and second is the test which is sles+extratests for ppc.
I will add the link with the working test later.
Updated by dgutu almost 8 years ago
- Status changed from In Progress to Resolved
- % Done changed from 90 to 100
Updated by dgutu almost 8 years ago
- Related to action #15036: snapper_cleanup dies with a reference to bsc#998360 which is closed added