OCHA-DAP / hdx-scraper-ophi / 33717501098
94%
main: 97%

LAST BUILD BRANCH: airflow
DEFAULT BRANCH: main
Ran
Jobs 1
Files 6
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 94.316%. Remained the same
33717501098

push

github

mcarans 
Fix false ERROR level and duplicate timestamp in task logs

Airflow 3 wraps every captured line of a task's output as one JSON
record per line (dag_id/run_id/task_id context plus the raw text under
"event") - this is how the Task SDK ships logs to the supervisor process
and isn't configurable away (confirmed: apache/airflow discussion #53006,
a maintainer response - "logs from the task process are sent as JSON to
the supervisor process and written to file"). It tags each record's
"level" purely by which stream the line came from, not its content:
stderr -> "error", stdout -> "info", regardless of whether the line was
actually an INFO message.

Our logger wrote to stderr, so every plain INFO message was mislabelled
"error" in the wrapper. Switched loguru's sink to stdout so the wrapper's
own level tag is now accurate, and dropped {level:...}/{time:...} from
our own format string (redundant now that the wrapper's outer "level"/
"timestamp" fields are correct) - verified live, a raw JSON record now
reads {"level":"info",...,"logger":"task.stdout"} correctly, with no
double-printed level or timestamp.

Also adds azure-setup/view-log.py: a small viewer that parses this
JSON-lines format back into conventional single-line log output
(`<timestamp> <LEVEL> <event>`) - the JSON-per-line format itself can't
be turned off since it's how Airflow 3 stores task logs at all, but
nothing stops reformatting it for reading, given it's still just a file.

531 of 563 relevant lines covered (94.32%)

0.94 hits per line

Jobs
ID Job ID Ran Files Coverage
1 tests - 33717501098.1 6
94.32
 GitHub Action Run
Source Files on build 33717501098