Prefect Reviews & Product Details

Processes that need to be run on an ad-hoc or on demand basis are very easy for our semi technical staff to use, and debugging is much easier with Prefect! Review collected by and hosted on G2.com.

Layout of the site changes frequently which is confusing to some of our team members, handling queues in an unhealthy state is difficult at times. Review collected by and hosted on G2.com.

I have used a variety of orchestrators in the past (such as Luigi, Airflow and prefect 1) and I wasn't really satisfied with any of them. When I tried prefect 2.0 I immediately fell in love with it and now I'm using it as my default orchestrator on my personal projects and I'm in the process of bringing it to the company I work for. What I enjoy the most is that you only need a decorator for defining a task and another one for defining a pipeline (called flow) and you don't need extra services or complexity to have a pipeline running. Also, it's extremely less verbose than Airflow. Review collected by and hosted on G2.com.

The only major problem I see with prefect is that it's still very new and that could come at the expense of having less support and development compared to other orchestrators. However, I'm willing to take the risk since I really like it. Review collected by and hosted on G2.com.

First of all, the speed of delivery by the Prefect team is simply amazing. They are always looking into improving the development and deployment aspects of the product. There is strong support through the Slack channel. I do like the rubber duck session to talk to the Prefect team members face-to-face'.

Please keep up the great work so far! Review collected by and hosted on G2.com.

On the flip side, due to the speed of the delivery of new features, it does require the end user of the product to closely follow the current state of the product in order to keep up with the latest. This is very necessary to avoid accumulating more tech debt. Just keep this in mind if you are part of a smaller organization. Review collected by and hosted on G2.com.

Since Prefect code tends to be just annotated regular code, testing is easy and intuitive, as one can just run the code as if it was regular Python and only testing it on Prefect infrastructure when necessary. There is also a large amount of flexibility in how deployment occurs due to the modularization of all the moving parts (flows, infrastructure, deployments). Review collected by and hosted on G2.com.

Since it's a relatively new tool, things change quickly and break quickly unless you're willing to lock into an older version. This introduces other problems, one of them the bugs that are present in the codebase. This is hard to avoid in such a new tool but it's a hassle anyway. The lack of a full cloud solution also causes some additional overhead, as workers (or agents in the past) are all the responsibility of the customer, even if they use the cloud solution. Review collected by and hosted on G2.com.

Everything works out of the box, you can even run your local flows and have them reflected in the prefect cloud dashboard. It's nice to worry so little about the orchestration and focus on the actual logic. The UI is awesome! Review collected by and hosted on G2.com.

Prefect 2 is nice however there is still some features that are in development which would have been ideal to have finished at launch such as built in Flow mapping. (Can still be accomplished with external libs at the moment) and a lot of the blocks need some polishing at the moment. It's still early for V2 so hopeful these will be resolved! Review collected by and hosted on G2.com.

Being able to write a flow just like a python script and then have it run as a job in a kubernetes cluster. Pydantic parameter validation is nice too. Review collected by and hosted on G2.com.

Upgrading from Prefect 1 to 2 was a lot of work on the infra side, and the new UI is a bit unintuitive. Onboarding a new user and getting them to understand the distinction between flows and deployments, and find what they're looking for in the UI wasn't straight forward.

There doesn't seem to be any way to stream flow run logs from the CLI which is annoying because we have flows that are triggered from a CI/CD pipeline and we want their execution to be blocking. I had to write a custom script to poll the API for flow run states.

We're currently experiencing intermittent issues with flow runs temporarily switching into a Failed state and then back into Running. So far have been unable to consistently reproduce or debug, but it breaks our integration tests regularly.

We're running flows in our AKS cluster with Dapr sidecars so that they can access databases and message queues. We've had issues configuring the sidecar to shut down properly at the end of a run, or when a run fails. I think this might be more of a Kubernetes issue than a Prefect one though. Review collected by and hosted on G2.com.

The hybrid architecture is game-changing for our business. Being able to schedule scripts in the cloud that are version controlled in GitHub and execute them on a virtual machine either on-prem or in the cloud is a huge benefit. Review collected by and hosted on G2.com.

It takes some time to get up to speed. It can be intimidating getting started because it's so flexible and you're really only limited by your imagination. Documentation is pretty solid and they regularly post content on YouTube. Review collected by and hosted on G2.com.

We've been using Prefect for a little while now and find it a very strong product. It operates our ELT processes on a robust schedule, and we also implemented Prefect for CI/CD. Bundling all workflows in one place is a great way to only rely on a single tool for orchestrating automated tasks and keeping track of the logs. Finally their team and community on slack are very responsive. Review collected by and hosted on G2.com.

The product has a steep learning curve and there is insufficient material in the form of a online learning platform or tutorials. Review collected by and hosted on G2.com.

The fact that I can just start with whatever Python code I have and then add a decorator is fantastic. It makes the development part of orchestration super easy for most developers to use and doesn't require a high degree of technical expertise to make functional code. Review collected by and hosted on G2.com.

The process for going from a simple POC to a fully functional Cloud (eg AWS ECS) platform is not documented very well and as such isn't that easy to get working. Review collected by and hosted on G2.com.

Pure python (the most likely known language for analyst), full range of technical possibilities (concurrency, retries, logs, notifications), easy to set up on Windows (no docker/linux necessary), very useful API to communicate with server, no disadvantages in comparison to Airflow found Review collected by and hosted on G2.com.

No button to "select all" after filtration (for delete action for example), no possibility to create authentication on server when using own server on linux + nginx Review collected by and hosted on G2.com.