Errors with large URLs

[Gamecock]

Describe your environment. Describe any aspect of your environment relevant to the problem, including your Python version, platform, version numbers of installed dependencies, information about your cloud hosting provider, etc. If you're reporting a problem with a specific version of a library in this repo, please check whether the problem has been fixed on master.
Python 3.8
Django 2.1
opencensus-ext-azure==1.0.4
opencensus-ext-django==0.7.2

App is running on a linux container on an Linux VM in Azure and intergrated with AzureAppInsights.

Steps to reproduce.
Describe exactly how to reproduce the error. Include a code sample if applicable.
Error occurs when reporting on requests a large URI.

What is the expected behavior?
What did you expect to see?
Request logged in AI, or at least dropped and app continues to run cleanly and error in AI.

What is the actual behavior?
What did you see instead?
DJANGO Logs:
Data drop 400: 100: Field 'url' on type 'RequestData' is too long. Expected: 2048 characters, Actual: 3701 {'iKey': , 'tags': {'ai.cloud.role': 'manage.py', 'ai.cloud.roleInstance': , 'ai.device.id': , 'ai.device.locale': 'en_US', 'ai.device.osVersion': '#1 SMP Tue Aug 25 17:23:54 UTC 2020', 'ai.device.type': 'Other', 'ai.internal.sdkVersion': 'py3.6.12:oc0.7.11:ext1.0.4', 'ai.operation.id': 'fcbe18bf6ca9036aa4546af171f3e877', 'ai.operation.name': 'GET /<my_url>/'}, 'time': '2020-12-15T17:58:36.498868Z', 'name': 'Microsoft.ApplicationInsights.Request', 'data': {'baseData': {'id': '116a0658b513bdb9', 'duration': '0.00:00:00.096', 'responseCode': '200', 'success': True, 'properties': {'request.name': 'GET /<my_url>/', 'request.url': 'https:///<my_url>/?=my_very_long_query_string', 'django.user.id': '90', 'django.user.name': '100044505'}, 'ver': 2, 'name': 'GET /<my_url>/', 'url': 'https:///<my_url>/?=my_very_long_query_string', 'source': None, 'measurements': None}, 'baseType': 'RequestData'}, 'ver': 1, 'sampleRate': None, 'seq': None, 'flags': None}.

We also see this repeating in the logs.

Queue is full. Dropping telemetry.
Queue is full. Dropping telemetry.
Queue is full. Dropping telemetry.
Queue is full. Dropping telemetry.
Queue is full. Dropping telemetry.
Queue is full. Dropping telemetry.

The app hangs up.

Additional context.
Add any other context about the problem here.
Stack Overflow Question:
https://stackoverflow.com/questions/65313480/django-opencensus-url-field-in-request-data-too-long

[lzchen]

This looks more like a restriction on django rather than an sdk issue correct?

[Gamecock]

I don't know where to look in the sdk, but I assume it's creating the RequestData object with default length fields, and then sending larger data to the object. I don't think I can override the object properties from my application, but I would love to be wrong.

[lzchen]

If you want to modify your telemetry before sending, you can change the envelope via telemetry processors.

[Gamecock]

@izchen I gave it my best shot, but the opencensus-ext-django documentation was not detailed enough for me.
I tried several things similar to:
` "handlers": {

    "azure": {
            
        "level": assert_env('DJANGO_LOG_LEVEL'),
        "class": "opencensus.ext.azure.log_exporter.AzureLogHandler",
        "instrumentation_key": assert_env('APPINSIGHTS_INSTRUMENTATIONKEY'),
        "telemetry_processor":"logging_callback_function"
         },`

without luck. Any hints are appreciated, or any hints to where RequestData is defined to just increase the field size.

[lzchen]

Telemetry_processors are an Azure specific concept, so it has nothing to do with opencensus-ext-django.

Judging by your error message, it appears that the url that is being sent in the RequestData data type is too long. There is no way currently to increase the limit for this field size.

The only thing you can do to circumvent this error is to make your url shorter by modifying the telemetry before it is sent to App insights.

def callback(envelope):
   envelope.data.baseData.url = "short_url"

handler = AzureLogHandler(instrumentation_key=<your-ikey>)
handler.add_telemetry_processor(callback)

[Gamecock]

@lzchen given the Azure Handlers are declared in the Django settings in JSON, I cannot figure out the syntax to add the telemetry processor.

[lzchen]

I would look in the django logging documentations to see if you can configure a handler to call a function.
If not, you can do either:

1. Move the logging instantiation and configuration into your actual Python code OR
2. After instantiating and configuring your logger through django settings, access your logger and handler programmatically and add the telemetry processor in your code. You can get the logger you are using by name and find the corresponding handler by logger.handlers.

[Gamecock]

Thanks

…

On Fri, Dec 18, 2020 at 12:56 PM Leighton Chen ***@***.***> wrote: I would look in the django logging documentations to see if you can configure a handler to call a function. If not, you can do either: 1. Move the logging instantiation and configuration into your actual Python code OR 2. After instantiating and configuring your logger through django settings, access your logger and handler programmatically and add the telemetry processor in your code. You can get the logger you are using by name and find the corresponding handler by logger.handlers. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#988 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABMSCSSCYUSDJ5YEA6OHBY3SVOJVFANCNFSM4U6PVKNQ> .

[lzchen]

Please let us know if your problem gets solved so we can close this ticket.

[Gamecock]

I tried a simple filter that just returns false, and logs keep going through.
I don't understand how getting a handler to the logger will help since I'm not explicitly logging the requests. I need a more detailed example.

[lzchen]

@Gamecock
Apologies, you have to add the handler to the logger that you are using.

import logging

from opencensus.ext.azure.log_exporter import AzureLogHandler

logger = logging.getLogger(__name__)

def callback(envelope):
   envelope.data.baseData.url = "short_url"

handler = AzureLogHandler(instrumentation_key=<your-ikey>)
handler.add_telemetry_processor(callback)

logger.addHandler(AzureLogHandler())
logger.warning('Hello, World!')

You are essentially doing what settings.py is doing (instantiating the Azure specific handler and attaching it to your logger instance), except manually through code.

Take a look at the examples for more examples on how to use the AzureLogHandler.

[Gamecock]

@izchen I understand, except it's not coming from the LOGGING Settings, it's coming from the OPENCENSUS settings:

`OPENCENSUS = {

'TRACE': {

    'SAMPLER': 'opencensus.trace.samplers.ProbabilitySampler(rate=1)',

    'EXPORTER': '''opencensus.ext.azure.trace_exporter.AzureExporter(
        service_name='mysite'
    )'''

    #Assumes Environmental Variable 'APPINSIGHTS_INSTRUMENTATIONKEY'
}

}
`
https://docs.microsoft.com/en-us/azure/azure-monitor/app/opencensus-python-request#tracking-django-applications

I don't know what to do next. I have a trivial sample I can share in github if that would help.

[lzchen]

Ahh my mistake. In that case are you able to do this in settings.py?

from opencensus.ext.azure.trace_exporter import AzureExporter
exporter = AzureExporter(<your-connection-string>)
exporter.add_telemetry_processor(callback)

OPENCENSUS = {
    'TRACE': {
        'SAMPLER': 'opencensus.trace.samplers.ProbabilitySampler(rate=1)',
        'EXPORTER': exporter,
    }
}

[Gamecock]

That worked, for anyone who follows along later the exported wants a vs .
@izchen do you want a PR to add the example to https://github.com/census-instrumentation/opencensus-python/tree/master/contrib/opencensus-ext-django?