Sentry Hook for Logrus
Sentry provides both self-hosted and hosted
solutions for exception tracking.
Both client and server are
open source.
Usage
Every sentry application defined on the server gets a different
DSN. In the example below replace
YOUR_DSN
with the one created for your application.
import (
"github.com/sirupsen/logrus"
"github.com/evalphobia/logrus_sentry"
)
func main() {
log := logrus.New()
hook, err := logrus_sentry.NewSentryHook(YOUR_DSN, []logrus.Level{
logrus.PanicLevel,
logrus.FatalLevel,
logrus.ErrorLevel,
})
if err == nil {
log.Hooks.Add(hook)
}
}
If you wish to initialize a SentryHook with tags, you can use the NewWithTagsSentryHook
constructor to provide default tags:
tags := map[string]string{
"site": "example.com",
}
levels := []logrus.Level{
logrus.PanicLevel,
logrus.FatalLevel,
logrus.ErrorLevel,
}
hook, err := logrus_sentry.NewWithTagsSentryHook(YOUR_DSN, tags, levels)
If you wish to initialize a SentryHook with an already initialized raven client, you can use
the NewWithClientSentryHook
constructor:
import (
"github.com/sirupsen/logrus"
"github.com/evalphobia/logrus_sentry"
"github.com/getsentry/raven-go"
)
func main() {
log := logrus.New()
client, err := raven.New(YOUR_DSN)
if err != nil {
log.Fatal(err)
}
hook, err := logrus_sentry.NewWithClientSentryHook(client, []logrus.Level{
logrus.PanicLevel,
logrus.FatalLevel,
logrus.ErrorLevel,
})
if err == nil {
log.Hooks.Add(hook)
}
}
hook, err := NewWithClientSentryHook(client, []logrus.Level{
logrus.ErrorLevel,
})
Special fields
Some logrus fields have a special meaning in this hook, and they will be especially processed by Sentry.
Field key | Description |
---|
event_id | Each logged event is identified by the event_id , which is hexadecimal string representing a UUID4 value. You can manually specify the identifier of a log event by supplying this field. The event_id string should be in one of the following UUID format: xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx and urn:uuid:xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx ) |
user_name | Name of the user who is in the context of the event |
user_email | Email of the user who is in the context of the event |
user_id | ID of the user who is in the context of the event |
user_ip | IP of the user who is in the context of the event |
server_name | Also known as hostname, it is the name of the server which is logging the event (hostname.example.com) |
tags | tags are raven.Tags struct from github.com/getsentry/raven-go and override default tags data |
fingerprint | fingerprint is an string array, that allows you to affect sentry's grouping of events as detailed in the sentry documentation |
logger | logger is the part of the application which is logging the event. In go this usually means setting it to the name of the package. |
http_request | http_request is the in-coming request(*http.Request). The detailed request data are sent to Sentry. |
Timeout
Timeout
is the time the sentry hook will wait for a response
from the sentry server.
If this time elapses with no response from
the server an error will be returned.
If Timeout
is set to 0 the SentryHook will not wait for a reply
and will assume a correct delivery.
The SentryHook has a default timeout of 100 milliseconds
when created
with a call to NewSentryHook
. This can be changed by assigning a value to the Timeout
field:
hook, _ := logrus_sentry.NewSentryHook(...)
hook.Timeout = 20*time.Second
Enabling Stacktraces
By default the hook will not send any stacktraces. However, this can be enabled
with:
hook, _ := logrus_sentry.NewSentryHook(...)
hook.StacktraceConfiguration.Enable = true
Subsequent calls to logger.Error
and above will create a stacktrace.
Other configuration options are:
StacktraceConfiguration.Level
the logrus level at which to start capturing stacktraces.StacktraceConfiguration.Skip
how many stack frames to skip before stacktrace starts recording.StacktraceConfiguration.Context
the number of lines to include around a stack frame for context.StacktraceConfiguration.InAppPrefixes
the prefixes that will be matched against the stack frame to identify it as in_appStacktraceConfiguration.IncludeErrorBreadcrumb
whether to create a breadcrumb with the full text of error