QA (public) » openQA Project (public)

I would like to add here that it would be very useful to be able to query, out of those jobs that are testing X incident, only their last run.
Also to be able to get only the failed last runs.

Yes, I suggest that we only yield the latest job for each

Following the suggestions I found this way to get the latest jobs (ids) that have failed and have the issue 14071

openqa-cli api --host http://localhost:9526 /jobs limit=1000 latest=1 |  jq '.[][] |  select(.settings |to_entries[]| .value | contains("14071")) | select(.result == "failed")  | .id'

Now this PR (https://github.com/os-autoinst/openQA/pull/3870) is only a draft and there are missing things but is the API approach to this feature

As the definition of "ISSUE" is something that SUSE QAM invented for their own internal use cases but there are other objects called "issue" within openQA itself the proposed API change would be a bad idea. I like the idea to introduce new API routes for good use cases but for now I suggest for this ticket to focus on work outside openQA source code. Do you consider the code snippet in #91542#note-9 enough to cover AC1? Feel welcome to crosscheck with users.

ilausuch wrote:

Following the suggestions I found this way to get the latest jobs (ids) that have failed and have the issue 14071

openqa-cli api --host http://localhost:9526 /jobs limit=1000 latest=1 |  jq '.[][] |  select(.settings |to_entries[]| .value | contains("14071")) | select(.result == "failed")  | .id'

this will easily fail on OSD which runs many jobs so the limit of 1000 won't be enough. A much more limited initial query result can be achieved by looking for the right build, e.g. to look for tests for incident ":19364:shim". one can do:

openqa-cli api --osd jobs build=":19364:shim" limit=1000 latest=1 | jq '.[][] | "\(.name): https://openqa.suse.de/t\(.id) (\(.result))"'

which gives output in format like:

"sle-15-SP2-Desktop-DVD-Incidents-x86_64-Build:19364:shim-qam-allpatterns@64bit: https://openqa.suse.de/t5916480 (softfailed)"
"sle-15-SP2-Desktop-DVD-Incidents-x86_64-Build:19364:shim-qam-gnome@64bit: https://openqa.suse.de/t5916481 (softfailed)"
…
"sle-15-SP1-Server-DVD-Incidents-Minimal-x86_64-Build:19364:shim-qam-minimal@uefi: https://openqa.suse.de/t5916833 (failed)"

this of course can be easily customized or filtered down further.

Thanks for the ideas

This returns the last failed jobs than has any issue 17 (this is an example, should be a complete issue value) and are done, and the result are failed

openqa-cli api --host http://localhost:9526 /jobs limit=1000 latest=1 |  jq  '.[][] |  select(.settings |to_entries[]| .value | contains("17")) | select(.result == "failed") | select(.state == "done" ) | {date: (.t_finished + "Z" | fromdate), id: .id, t_finished: .t_finished}' | jq --slurp 'group_by(.id) | .[] | last'

{
  "date": 1583845259,
  "id": 3976352,
  "t_finished": "2020-03-10T13:00:59"
}
{
  "date": 1583851807,
  "id": 3976455,
  "t_finished": "2020-03-10T14:50:07"
}
{
  "date": 1583852617,
  "id": 3976457,
  "t_finished": "2020-03-10T15:03:37"
}
{
  "date": 1583860604,
  "id": 3976572,
  "t_finished": "2020-03-10T17:16:44"
}
{
  "date": 1583861145,
  "id": 3976730,
  "t_finished": "2020-03-10T17:25:45"
}

I'm curious. if you do select(.state == "done" ), why not use that as a parameter for the API call?
/jobs limit=1000 latest=1 state=done

Good point. Will do. The reason is because I was using a output.json file to no interact with the API all the time

I had a chat with @geor and some things came up

• He need to choose more than one options for result 'failed' but also 'parallel_failed' for now
• About what @tinita mentioned, could in the openqa_cli command line could we specify multiple result values (OR)?
• Remove the date to epoch thing is not necessary

ilausuch wrote:

Good point. Will do. The reason is because I was using a output.json file to no interact with the API all the time

And that makes sense if multiple requests are done based on the same data set.

ilausuch wrote:

• About what @tinita mentioned, could in the openqa_cli command line could we specify multiple result values (OR)?

I don't think this is necessary but I think it's not necessary. As you already did you can gather the data from openQA in a more generic query and filter that down with e.g. jq locally

Well,
Based on the comments this could be a good approach

openqa-cli api --host http://localhost:9526 /jobs limit=1000 latest=1 > output.json

cat output.json |  jq  '.[][] |  select(.settings |to_entries[]| .value | contains("17")) | select(.result == "failed" or .result == "parallel_failed") | select(.state == "done" ) | {id: .id, t_finished: .t_finished}' | jq --slurp 'group_by(.id) | .[] | last'

Based on the solution for issues this solves the same problem for packages

cat output.openseuse.json | jq  '.[][] |  select(.settings |to_entries[]| .value | contains("yast2-network")) | select(.result == "failed" or .result == "parallel_failed") | select(.state == "done" ) | {id: .id, t_finished: .t_finished, group: .group, name: .name}

Where results are:

...
{
  "id": 1725317,
  "t_finished": "2021-05-03T10:05:10",
  "group": "openSUSE Tumbleweed",
  "name": "microos-Tumbleweed-MicroOS-Image-ContainerHost-x86_64-Build20210502-container-host-old2microosnext@64bit"
}
{
  "id": 1725471,
  "t_finished": "2021-05-03T10:29:46",
  "group": null,
  "name": "opensuse-Tumbleweed-DVD-x86_64-ltp_kernel_misc:investigate:retry@64bit"
}

Additionally the group of packages can be included, such as "REPO_OSS_SOURCE_PACKAGES"

cat output.openseuse.json | jq  '.[][] | select(.settings | to_entries[] | .key | contains("REPO_OSS_SOURCE_PACKAGES")) | select(.settings |to_entries[]| .value | contains("yast2-network")) | select(.result == "failed" or .result == "parallel_failed") | select(.state == "done" ) | {id: .id, t_finished: .t_finished, group: .group, name: .name}'

With George we identified two different scenarios for the (issues) AC

1. Single incident jobs: Where each build is related only with one incident
2. Aggregated jobs: Where each build can contains more than one incident.

We extract information using tree API calls

openqa-cli api --host http://openqa.suse.de /jobs limit=5000 result="passed" latest=1 >output.passed.json
openqa-cli api --host http://openqa.suse.de /jobs limit=5000 result="failed" latest=1 >output.failed.json
openqa-cli api --host http://openqa.suse.de /jobs limit=5000 result="softfailed" latest=1 >output.softfailed.json

And we created an initial script for each scenario

1. Single jobs

#!/bin/bash
#Usage: $0 <issue>

filter_group="select(.group != null) | select(.group | contains(\"Incident\"))"
issue="select(.settings | to_entries[] | .value | contains(\"$1\"))"

failed=$(cat output.failed.json | jq ".[][] | $filter_group | $issue | .name " | sort | uniq)

for name in $failed; do
  echo $name
  failed_ts=$(cat output.failed.json | jq ".[][] | $filter_group | $issue | select(.name==$name)" | jq --slurp 'group_by(.name) | .[] | last | .t_finished+"Z" | fromdate')
  job_id=$(cat output.failed.json | jq ".[][] | $filter_group | $issue | select(.name==$name)" | jq --slurp 'group_by(.name) | .[] | last | .id')
  passed_ts=$(cat output.passed.json | jq ".[][] | $filter_group | $issue | select(.name==$name)" | jq --slurp 'group_by(.name) | .[] | last | .t_finished+"Z" | fromdate')
  softfailed_ts=$(cat output.softfailed.json | jq ".[][] | $filter_group | $issue | select(.name==$name)" | jq --slurp 'group_by(.name) | .[] | last | .t_finished+"Z" | fromdate')

  [[ ! $passed_ts ]] && passed_ts=0
  [[ ! $softfailed_ts ]] && softfailed_ts=0

  [[ $failed_ts -gt $passed_ts ]] && [[ $failed_ts -gt $softfailed_ts ]] && echo "Not passed $job_id"
done

1. Aggregated jobs

#!/bin/bash

#Usage: $0 <issue> <group>

add_hash='.hash=(.settings.DISTRI+"-"+.settings.VERSION+"-"+.settings.FLAVOR+"-"+.settings.TEST+"-"+.settings.ARCH+"-"+.settings.MACHINE)'
filter_group="select(.group != null) | select(.group | contains(\"$2\"))"
issue="select(.settings | to_entries[] | .value | contains(\"$1\"))"

failed=$(cat output.failed.json | jq ".[][] | $filter_group | $issue | $add_hash | .hash")

for hash in $failed; do
  echo $hash
  failed_ts=$(cat output.failed.json | jq ".[][] | $filter_group | $issue | $add_hash | select(.hash==$hash)" | jq --slurp 'group_by(.hash) | .[] | last | .t_finished+"Z" | fromdate')
  job_id=$(cat output.failed.json | jq ".[][] | $filter_group | $issue | $add_hash | select(.hash==$hash)" | jq --slurp 'group_by(.hash) | .[] | last | .id')
  passed_ts=$(cat output.passed.json | jq ".[][] | $filter_group | $issue | $add_hash | select(.hash==$hash)" | jq --slurp 'group_by(.hash) | .[] | last | .t_finished+"Z" | fromdate')

  [[ ! $passed_ts ]] || [[ $failed_ts -gt $passed_ts ]] && echo "Not passed $job_id"
done

The aggregated script is already confirm, but the single one is pendent.
Also these are initial scripts than can be optimized.

Talking with George, we think that the second AC should be an other ticket because of

• It's a different problem with different solutions
• To solve that is needed to access to the comments in the tests source code

ilausuch wrote:

Talking with George, we think that the second AC should be an other ticket because of

• It's a different problem with different solutions
• To solve that is needed to access to the comments in the tests source code

I feel like you went way overboard here. The AC was "example API call" and you provide two nice looking scripts :-D

For the second one I'd suggest the search API route which can filter by Package: if that's what you want here, for instance:

cat https://openqa.opensuse.org/api/v1/experimental/search?q=Package:

FYI as a follow-up I suggested a workshop topic about openQA API. And I think @bzoltan @geor @ilausuch one of you guys should file some follow-up tickets with concrete user stories (think about where the API would be used e.g. in scripts, curl, smelt, openQA UX)

Workshop has been conducted, recording https://youtu.be/EfXZKbQS-Kg