Skip to content

On Windows NUC 32GB system memory - zombie nodes not released #164

@LeeBamberTGC

Description

@LeeBamberTGC

I am running a project with 125 features, just one agent and keeping performance monitor open. I notice sometimes the agent can get into trouble and system memory usage starts going up until it hits 100% and starts thrashing the swap file. This is okay for a while, but then the OS starts to stutter. I found that pausing the build, exiting AUTOFORGE in the PowerShell with CTRL+C and then using the command "taskkill /F /IM node.exe" I am able to kill off all the zombie nodeJS runtime processes that linger and restoring my physical memory available. Restarting AUTOFORGE then works smoothly until more zombie nodes are collected.

I suspect that agents are starting these nodes, but not tracking and deleting the "child" nodes when the feature is completed. I just went ahead and purchased a new 64GB SODIMM kit before I realised that it might be a bug rather than a feature :) Still, my new purchase will allow me to run more agents in parallel :)

Feel free to ask for more info and shots. I may also tinker with the repo at some point and see if I can apply a fix. In the meantime, here is the offending process in my process monitor:

Image

P.S. EXCELLENT TOOL!!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions