I have a flask API that handles ML models and I want to get metrics based on each model but the way we grab each model is based on the path. ex: https://api.someapi.com/cmd/<model-name> is it possible to grab metrics per model name? I was reading the docs and I guess we could hack something together with labels but that doesn't seem like the right solution. Has anyone run into this problem? I am using prometheus-flask-exporter.
labels is always the right solution in Prometheus' to differentiate between otherwise identical metrics.
The "trick" is in identifying the unique metrics and the way to determine these is to ask "Will we ever want to (dynamically) query metrics by all or by some label values?". Corollary: what measurements of your models will you want to compare across models?
Without knowing more about your solution, you may have metrics corresponding to model invocations, failures etc. and you would differentiate between e.g. model invocations by applying at least one label.
In your case, perhaps a label key would be model_name and you'd use the path parameter <model-name> as the value.
It's unclear whether your paths are statically or dynamically generated but it makes no difference to the metrics and labels; just how you apply them.
Related
When project gets bigger I need to see all events which happens during model or project lifecycle.
EDIT
I don't want Django to generate some diagram, I want to do it manually.
I want to see (on graph or chart) all consequences of actions like creating, updating, saving etc.
For example, when I create a Property object, there is a post_save signal which creates PropertyProfile and PropertyProfile has overriden save method where I calculate price of the property before super().save(..).
Which tool should I use to visualize such events so I can see what will happen and when will it happen?
I think that sequence diagram or flowchart is probably not enough because I can't visualize all such information but maybe I'm wrong, not good at UML.
to achieve this ,you first need to generate data; write log or define a model, create instance when signal fired;
to visualize such data .It almostly depends on how many data there.
ELK is suitable if you have large large amount data;otherwise , Echarts maybe is enough
I am trying to do some reporting on page views on a site and the results are being listed like the following:
www.example.com/directory/ - 100 views
www.example.com/directory/?id=123456 - 10 views
www.example.com/directory/?id=987654 - 5 views
What filter do I need to create to views the results as:
www.example.com/directory/ - 100 views
www.example.com/directory/?id=* - 15 views
Thanks in advance
Yes, getting historical grouped together is going to mean using something like Google Docs, Excel, Tableau Software, Analytics Canvas, etc.
Moving forward...
One of the simplest ways of keeping things grouped in GA is to set up an advanced profile filter. You'll want to use this with a new profile; keeping a "raw" or "empty" profile is highly advisable for when you actually want to look at those individual URLs.
That said, here's a filter pattern that should work for you:
Go to Admin > Filters (under the View Column)
+ New Filter > Create new Filter > Name it
Filter Type = Custom filter > Advanced
Here's the pattern:
Field A: www\.example\.com\/directory\/\?id=.+
Output To: www\.example\.com\/directory\/\?id=\*
Another way to aggregate the same URI with multiple query strings is to change the primary dimension to 'Page Title' under Behavior > Site Content > All Pages.
The best way to do this for your historical data is unfortunately in an excel pivot table. You can get in in the UI, but only by creating a custom report and searching for very specific directories.
Check out the documentation on excluding query strings in your GA profile. Maybe create a new profile and write an advanced rule to rewrite all "id" pages to "/directory/product-page".
A totally different approach is to use custom variables or custom dimensions and to stop looking in the normal "Behavior" reports section (used to be called "Content" in GA) – custom dims are available using Google Analytics Universal Analytics only, which means starting a new web property and possibly running both code snippets concurrently (totally safe to do).
Personally I find custom dimensions a bit easier to work with than custom variables, and I generally think that it's a good idea to start exploring the new Google Analytics.
The nice thing about either of these approaches is that you can still keep the full page path date in the same profile as your custom dimension / variables information; it'll stay in the Behavior section where it belongs with all the other page paths.
Where I'm going with this...
You can create a new dimension such as "page type" and then call it "products", "posts", "articles", or whatever these id #s represent in this /directory/; then you can look at metrics across the dimension like pageviews, time on page, etc. by page type.
You can even create other dimensions to help describe them in more detail, such as breaking down blog posts or products into their different categories; i.e. hierarchical dimensions. Once you start using this kind of thing you may wonder what you ever did without it!
I think it's fair that I stop this answer now since it's not about how to set up custom variables or custom dimensions; those links should get you started (it's really not difficult).
Note: You can use php to fill in the dimension information in the GA tracking snippet dynamically based on the page that is being viewed (again, that's another question).
I have a page where I'm displaying the results of a queryset to the user.
What i'd like to do is allow the user to click on a link in order to apply a filter.
Currently what I do is have the links pass "get" parameters to the page in order to apply filters. The filters can be references to other models or custom filters (e.g. an unassigned filter)
In order to provide a decent user experience the implementation needs to do a few things
in the view:
check that the filter parameter passed is valid
check what type of filter it is (based on other models or a custom filter) in order to apply the correct condition to the queryset
(optional) a way to make the filters cumulative (i.e. you can keep adding filters)
in the Template:
display the correct resultset based on the filter choosen
when displaying the filters, recognize which filter we have applied so that the current applied filter is displayed as text not a hyperlink.
I'm thinking this must be common enough that someone must have like a design pattern or best practice figured out for this other than the obvious whack of if/else statements in the view and the template.
is there?
I find the way the Django admin handles this kind of functionality a great pattern. If you're not familiar, check out the list_filter option in the admin. It's similar to what you're describing, but yours is a bit more generic. Perhaps this will help you ponder some ideas?
First, for the actual querystring chunk, you're simply passing the Django-ORM lookup key and value pair. e.g., ?sites__id__exact=1, tags__in=words, etc. Since you want to allow for cross-model lookups, you'd need to provide another parts in the string to include the model name, not too tough.
For checking if the filter is valid, you can simply ensure that the model/field lookup is valid. By splitting the parts of each QS chunk, you can identify the model, the fieldname, the lookup, and the value. Then, use Django's built-in functionality to validate that fieldname exists on model. You can do this with ForeignKey's too. Here's how Django does it
You can keep adding filters pretty easily to this. You'll be providing your view and the form that's displaying these filters with some context, so it'll persist and re-populate for the user. Also, you could just as easily persist the query string. Basically, you'd have the same read / parsing functionality here at all times, nothing really different.
I think the keys are automating and keeping it as DRY as possible. Don't succumb to a bunch of if statements. It's really easy to pass these lookups into the ORM, safely too, and it's really easy to catch bad lookups and provide the user with a meaningful error message.
I hope that helps you on your path! :)
I am developing an application in Django and I am curious on how I can go about adding a model such that only 1 row is only ever present (i.e. Singleton).
As an example, I'd like to maintain a set of boolean flags of the application i'm running as to whether: it's on or off (so I can manually turn it on or off, perhaps even per module).
I can't see any part of the docs explaining a good way to go about setting this up.
Any suggestions?
Not sure from you explanation in what context you require this but I have a model which holds a number of key/value pairs used in validator checks and other things. The keys are all needed by each implementation of the project but the values will differ between projects. The values should be maintainable by an admin user. The values usually do not need to change very much once set. Given that, I decided to put them in a model. It is a bit weird but simple enough.
You should be able to limit write access to the model to the one row for either your app or your users through your code.
only ever reference the first row in the QuerySet
row = MyVariables.objects.all()[0]
Test if there are rows first. if you think there might accidentally be more than one record then make sure it is ordered (but that should never happen if you did (1) correctly.
There are a couple of apps already dealing with this, check out http://djangopackages.com/grids/g/live-setting/
I'm also a bit confused on your goal, but I'd recommend looking at the Model Instance section of the docs. You should probably look at customizing the validation or cleaning of the model.
If your goal is to only have 1 row flagged in the table for your model: during the validation you can run a query to see if any other row is flagged, and update them to be not flagged. (or delete them).
This question Unique BooleanField value in Django may be helpful.
Hi everyone I have a few questions about the django admin.
First the relevant details. I currently have Client, Printer, Cartridge, and Order models.
The Printer model has a ManyToManyField to the Cartridge model, which would allow you to select all the cartridges that can be used with that printer.
The Cliente has a ManyToManyField to the printers which they own.
1) I want to create an Order through the Django admin which lets your specify the Client, a dicount, and multiple cartridges through a ManyToManyField. This is getting kinda tricky because I have to do it through another table that specifies whether it's a new Cartridge or a refill.
2) I want the admin to filters the Cartridges to only show the ones that belong to the printers that they own.
3) Also I would like to have a field that holds the total price of their order, but it should calculate it based on how many cartridges they have added to the order. I don't know if this should be done by adding more of the same cartridge to the order or by having another field in the related table that specifies the quantity.
Can this be done in the admin or do I need to use a form? And if so how would I go about adding this to the admin? It seems difficult and probably something I will have to do in multiple parts since in order to filter the list of cartridges I have to know the client beforehand.
As far as I can see, no, it's not really possible. The development version has some methods for limiting foreign keys, but it doesn't seem to me that limiting based on the customer is possible, since it depends on separate foreign keys.
The best suggestion, if you're really bent on doing it in the admin form, would be to use Javascript to do it. You would still have to make AJAX calls to get lists of what printers customers had and what cartridges to show based on that, but it could be done. You would just specify the JS files to load with the Media class.
But I think that's more work than it's worth. The easiest way I would see to do it would be with Form Wizards. That way, you'd have a step to select the customer so on the next step you know what cartridges to show.
Hope that helps!
I've worked similar problems, and have come to the conclusion that in many cases like this, it's really better to write your own administration interface using forms than it is to try and shoehorn functionality into the admin which is not intended to be there.
As far as 3) goes, it depends on what your product base looks like. If you're likely to have customers ordering 50 identical widgets, you probably do want a quantity field. If customers are more likely to be ordering 2 widgets, one in red, one in blue, add each item separately to the manytomany field and group them in your order interface.