Project

General

Profile

action #25982

Updated by okurz about 6 years ago

## Observation 

 openQA test in scenario sle-15-Installer-DVD-s390x-skip_registration+workaround_modules@zkvm fails in 
 [bootloader_zkvm](https://openqa.suse.de/tests/1209036/modules/bootloader_zkvm/steps/25) 

 Message in autoinst-log.txt: 

 ``` 
 no/incomplete answer. 
 Please make sure your installation medium is available. 
 ``` 


 ## Reproducible 

 Fails since (at least) Build [300.1](https://openqa.suse.de/tests/1209036) (current job) 


 ## Acceptance criteria 

 **AC1:** DONE: <del>get get an idea about the root cause</del> cause 
 **AC2:** DONE: <del>gather gather statistics - it's pretty sporadic</del> 
 sporadic 

 **optional:** 

 **AC3:** Come up with a fix or a workaround to mitigate the issue if easy, otherwise create followup ticket for implementation 


 ## Expected result 

 Last good: [299.1](https://openqa.suse.de/tests/1208375) (or more recent) 

 ## Problem 

 Happens in about 5/100 cases (see comments) and most likely for the first tests in a new build, shortly after a new build hits openQA. Retriggering helps here. 

 * **H1:** Jobs are scheduled before the build was actually completely synced to osd -> **E1-1** Check logs of osd:/var/log/openqa_rsync.log and the jobs log to see if the job is triggered before syncing is complete 
 * **H2:** Something about caching over network? 
 * **H3:** Many jobs starting at the same time cause an overload causing the installer to get a negative reply trying to access the FTP server -> 
  * **E3-1** Trigger many jobs in parallel for an existing build, e.g. the previous build when the test queue for s390x is empty 
  * **E3-2** Check if there are any logs on osd regarding the ftp server to see if there was a deny or apache logs? 


 ## Further details 

 Always latest result in this scenario: [latest](https://openqa.suse.de/tests/latest?distri=sle&flavor=Installer-DVD&arch=s390x&test=skip_registration%2Bworkaround_modules&version=15&machine=zkvm) 


 ## Workaround 
 Retrigger jobs. They should be able to find the repo after retriggering.

Back