structlog-sentry

Based on https://gist.github.com/hynek/a1f3f92d57071ebc5b91

Installation

Install the package with pip:

pip install structlog-sentry

Usage

This module is intended to be used with structlog like this:

import sentry_sdk
import structlog
from structlog_sentry import SentryProcessor


sentry_sdk.init()  # pass dsn in argument or via SENTRY_DSN env variable

structlog.configure(
    processors=[
        structlog.stdlib.add_logger_name,  # optional, must be placed before SentryProcessor()
        structlog.stdlib.add_log_level,  # required before SentryProcessor()
        SentryProcessor(event_level=logging.ERROR),
    ],
    logger_factory=structlog.stdlib.LoggerFactory(),
    wrapper_class=structlog.stdlib.BoundLogger,
)


log = structlog.get_logger()

Do not forget to add the structlog.stdlib.add_log_level and optionally the structlog.stdlib.add_logger_name processors before SentryProcessor. The SentryProcessor class takes the following arguments:

• level Events of this or higher levels will be reported as Sentry breadcrumbs. Default is logging.INFO.
• event_level Events of this or higher levels will be reported to Sentry as events. Default is logging.WARNING.
• active A flag to make this processor enabled/disabled.
• as_context Send event_dict as extra info to Sentry. Default is True.
• ignore_breadcrumb_data A list of data keys that will be excluded from breadcrumb data. Defaults to keys which are already sent separately, i.e. level, logger, event and timestamp. All other data in event_dict will be sent as breadcrumb data.
• tag_keys A list of keys. If any if these keys appear in event_dict, the key and its corresponding value in event_dict will be used as Sentry event tags. use "__all__" to report all key/value pairs of event as tags.
• ignore_loggers A list of logger names to ignore any events from.
• verbose Report the action taken by the logger in the event_dict. Default is False.
• scope Optionally specify sentry_sdk.Client (in structlog-sentry<2.2 this corresponds to hub: sentry_sdk.Hub).

Now events are automatically captured by Sentry with log.error():

try:
    1/0
except ZeroDivisionError:
    log.error("zero divsiion")

try:
    resp = requests.get(f"https://api.example.com/users/{user_id}/")
    resp.raise_for_status()
except RequestException:
    log.error("request error", user_id=user_id)

This won't automatically collect sys.exc_info() along with the message, if you want to enable this behavior, just pass exc_info=True.

When you want to use structlog's built-in format_exc_info processor, make that the SentryProcessor comes before format_exc_info! Otherwise, the SentryProcessor won't have an exc_info to work with, because it's removed from the event by format_exc_info.

Logging calls with no sys.exc_info() are also automatically captured by Sentry either as breadcrumbs (if configured by the level argument) or as events:

log.info("info message", scope="accounts")
log.warning("warning message", scope="invoices")
log.error("error message", scope="products")

If you do not want to forward a specific logs into Sentry, you can pass the sentry_skip=True optional argument to logger methods, like this:

log.error("error message", sentry_skip=True)

Sentry Tags

You can set some or all of key/value pairs of structlog event_dict as sentry tags:

structlog.configure(
    processors=[
        structlog.stdlib.add_logger_name,
        structlog.stdlib.add_log_level,
        SentryProcessor(event_level=logging.ERROR, tag_keys=["city", "timezone"]),
    ],...
)

log.error("error message", city="Tehran", timezone="UTC+3:30", movie_title="Some title")

this will report the error and the sentry event will have city and timezone tags. If you want to have all event data as tags, create the SentryProcessor with tag_keys="__all__".

structlog.configure(
    processors=[
        structlog.stdlib.add_logger_name,
        structlog.stdlib.add_log_level,
        SentryProcessor(event_level=logging.ERROR, tag_keys="__all__"),
    ],...
)

Skip Context

By default SentryProcessor will send event_dict key/value pairs as contextual info to sentry. Sometimes you may want to skip this, specially when sending the event_dict as sentry tags:

structlog.configure(
    processors=[
        structlog.stdlib.add_logger_name,
        structlog.stdlib.add_log_level,
        SentryProcessor(event_level=logging.ERROR, as_context=False, tag_keys="__all__"),
    ],...
)

Ignore specific loggers

If you want to ignore specific loggers from being processed by the SentryProcessor just pass a list of loggers when instantiating the processor:

structlog.configure(
    processors=[
        structlog.stdlib.add_logger_name,
        structlog.stdlib.add_log_level,
        SentryProcessor(event_level=logging.ERROR, ignore_loggers=["some.logger"]),
    ],...
)

Logging as JSON

If you want to configure structlog to format the output as JSON (maybe for elk-stack) you have to disable standard logging integration in Sentry SDK by passing the LoggingIntegration(event_level=None, level=None) instance to sentry_sdk.init method. This prevents duplication of an event reported to sentry:

from sentry_sdk.integrations.logging import LoggingIntegration


INTEGRATIONS = [
    # ... other integrations
    LoggingIntegration(event_level=None, level=None),
]

sentry_sdk.init(integrations=INTEGRATIONS)

This integration tells sentry_sdk to ignore standard logging and captures the events manually.

Testing

To run all tests:

tox

Contributing

Create a merge request and tag @kiwicom/platform for review.