[GSoC] Mass Commands

[nemesifier]

Official thread for this GSoC idea, ask questions and announce your interest here.

[Srinath0916]

Hey @nemesifier @pandafy ,

I've been analyzing the Mass Commands project and went through the current command execution implementation, specifically the Command model in connection/base/models.py, the launch_command task, and the change_group admin action pattern.

Based on this, I'm planning to implement a MassCommand model that uses bulk_create() within transaction.atomic() to create individual Command objects for each target device, letting them execute in parallel through the existing launch_command task.

I have a few questions:

1. Connection Pool Limits: For operations targeting 1000+ devices, should the orchestrator implement chunking (e.g., create commands in batches of 500) to prevent exhausting SSH/VPN connections, or should we rely on Celery's concurrency limits?

2. WebSocket Updates: To keep the browser responsive when many devices complete simultaneously, should I batch WebSocket updates (e.g., every 1-2 seconds) rather than sending immediate per-device updates?

3. Retry Audit Trail: When retrying failed devices, should we create new Command objects to preserve history, or update existing ones?

I'm preparing a detailed proposal and would appreciate guidance on these decisions.

[nemesifier]

1 and 2) it's hard to respond to these questions now as it could be risking premature optimization. In general I would keep it simple, especially in the first phase. The config options of celery allow us to control concurrency, while for websockets, I am not worried really. There's not that many concurrent workers which would allow overload the system significantly. If we ever run into performance issues we can deal with it later, I am not a fan of premature optimization. UX in this project is much more important.

Regarding 3), in the firmware upgrader module we reuse the same object so I'd stay consistent.

[Srinath0916]

Thanks for clarifying!, understood, I'll keep it simple and focus on UX more importantly.
I'll plan to follow the firmware upgrader pattern for retry.

[dee077]

Hi @nemesifier,

I would like to express my interest in working on this project this year. I feel I have already worked on similar tasks mentioned in the requirements during the last term’s project. Currently, I am reviewing the single command implementation and its flow to better understand its current limitations and how the mass command feature could extend it.

While exploring the idea, I had a couple of questions:

1. For the UI/UX part of this project, will contributors be expected to submit Figma designs that the UI/UX designers will later refine similar to last year, or will designers implement the UI directly from scratch this time?

2. For testing the mass command functionality, will it be necessary to flash OpenWrt on a physical router, or would a virtualized router setup, e.g., OpenWrt running in a VM, be acceptable?

   I am asking because VM-based routers can have some limitations, for example:

   • Limited or simulated network interfaces
   • Lack of real wireless interfaces and signal data
   • Some networking features (like hardware-specific drivers) are not being available
   • Potential differences in performance compared to real devices

As I continue exploring the implementation and understand it better, I may encounter additional challenges. I will post further questions here as they come up.

Thanks.

[nemesifier]

1. Including UX mockups is always a good exercise that forces the applicant to think thoroughly and propose a solution that matches the expected outcomes. You can count on the fact that we'll refine the UX mockups.

2. VMs are fine for testing commands, the actual commands being tested do not matter much, we can test any command that works.

[Yashika0724]

Hi @nemesifier, I'm Yashika and I'd like to express my interest in working on Mass Commands for GSoC 2026. I've been contributing to openwisp-controller recently (#1268, #1269, #1270) and have been studying the existing Command model and launch_command task to understand the foundation this feature builds on. I plan to include detailed UX mockups in my proposal. Looking forward to contributing!

[dee077]

Hi @nemesifier,

I was looking for other places where a similar feature is implemented to get some UI inspiration and understand the flow design. One of the most widely used examples I found is the AWS Systems Manager – Run Command feature. In that UI, users can select instances from a list, provide a textarea to enter the command, and then view the execution results once the command runs.

What I am thinking is to add a new page under the Configuration tab in the side menu, with a new section called Run Command. We could first show a list of devices (initially limited to around 5–6 devices per page) with filters such as organization, location, status, etc., and similar to AWS, after selecting devices, the UI would show a textarea where the command can be entered along with working dir and timeout.

Once the command is executed, we can add an inline tab like Recent Commands, which we already have for single command, and redirect to this tab to display the results of the execution. Let me know if this idea is worth moving forward with, or if you have inspiration from other systems that you think would be better to follow.

If you agree with this direction, I will move forward with creating some mockups for the UI.