Rykarix
commented
2 weeks ago Additional resources:
Lookup structlog if you want a library that can handle the json aspect of things for you.
Otherwise just write a file handler for it. I created one similar for another logging project:
# src/logs/formatters/json.py
"""Modified JSON Formatter inspired by Eric Moss.
github: https://github.com/erok77/esdemo
yt logging vid: https://youtu.be/_hhO56anumg
"""
import json
import logging
from datetime import UTC, datetime
def remove_none(d: dict) -> dict:
"""Removes None values from nested object."""
try:
if not isinstance(d, dict | list):
return d
if isinstance(d, list):
return [v for v in (remove_none(v) for v in d) if v]
return {k: v for k, v in ((k, remove_none(v)) for k, v in d.items()) if v}
except Exception:
logging.exception("Issue stripping None vals from nested object")
def format_traceback(record: logging.LogRecord) -> list:
"""Converts traceback into JSON-friendly format."""
structured_traceback = []
tb = record.exc_info[2]
while tb:
frame = tb.tb_frame
structured_traceback.append(
{
"filename": frame.f_code.co_filename,
"name": frame.f_code.co_name,
"lineno": tb.tb_lineno,
"locals": dict(frame.f_locals),
},
)
tb = tb.tb_next
return structured_traceback
class JsonFormatter(logging.Formatter):
"""JsonFormatter : extends logging.Formatter."""
def get_exc_fields(self, record: logging.LogRecord) -> dict:
"""Returns execution/exception info for log record."""
if record.exc_info:
return {
"exc_type": record.exc_info[0].__name__,
"exc_message": str(record.exc_info[1]),
"stack_trace": format_traceback(record),
}
return {"exc_info": record.exc_text}
def format(
self,
record: logging.LogRecord,
*args: str,
**kw: dict,
) -> str:
"""Takes log record and returns JSON Structured Logging Format."""
json_log_object = {
"@timestamp": datetime.now(UTC).isoformat(),
"error": {
"stack_trace": self.get_exc_fields(record),
},
"event": {
"created": datetime.fromtimestamp(record.created, tz=UTC).isoformat(),
"module": record.module,
},
"log": {
"level": record.levelname,
"logger": record.name,
"origin": {
"file": {
"line": record.lineno,
"name": record.filename,
"fullpath": record.pathname,
},
"function": record.funcName,
},
},
"message": record.getMessage(),
"process": {
"pid": record.process,
"name": record.name,
"thread": {
"id": record.thread,
"name": record.threadName,
},
},
}
return json.dumps(remove_none(json_log_object), indent=2)
def test_trace_3() -> None:
"""Raise an exception to trace."""
msg = "An error occurred in level 3"
raise ValueError(msg)
def test_trace_2() -> None:
"""Raise an exception to trace."""
try:
test_trace_3()
except ValueError as e:
msg = "An error occurred in level 2"
raise RuntimeError(msg) from e
def test_trace_1() -> None:
"""Raise an exception to trace."""
try:
test_trace_2()
except RuntimeError:
logger.exception("An error occurred in level 1")
if __name__ == "__main__":
logger = logging.getLogger("TestLogger")
handler = logging.StreamHandler()
handler.setFormatter(JsonFormatter())
logger.addHandler(handler)
logger.setLevel(logging.ERROR)
test_trace_1()
Mage version
Latest
Describe the bug
Issue 1
does not save logs as json files. It saves them as log files.
Issue 2
The output of these "json logs" is broken. The following is not valid json format:
Issue 2.5
Weirdly enough, some logs shown in the docker-compose terminal show correctly formatted json output
Issue 3
There is no clear method of setting the server_logs folder apart from the pipeline_logs folder.
Issue 4
Errors are not being saved to logs properly.
For instance, I can see in the docker-compose up terminal that
Yet this doesn't appear in logging.
Issue 5
There's no clear instructions, and probably even method, to extend your DX_LOGGER so that we can add our own streaming handlers on building a dockerfile.
To reproduce
Just set
ENV SERVER_LOGGING_FORMAT=jsonand everythingIn fact, here's my whole dockerfile:
Expected behavior
Issue 1
Logs should be saved as .json not .log
Issue 2
Log outputs should have valid json formatting
Issue 3
I expect non-pipeline related logging to be saved seperately to pipeline-related logging. For instance, if there's an error about user permissions or if there's a warning about something being insecure - that's related to mage as an application & its deployment. All of these logs should have a proper funnel
Issue 4
I expect all errors and warnings to be logged correctly based on
ENV SERVER_VERBOSITY=WARNINGIssue 5
There should be a streamhandler that can connect to various services like elastic search, however since this is somewhat a feature request I'll add that into a seperate ticket
Screenshots
No response
Operating system
See "To reproduce"
Additional context
See "To reproduce"