Thanks to visit codestin.com
Credit goes to github.com

Skip to content

Conversation

@pkova
Copy link
Collaborator

@pkova pkova commented Aug 25, 2024

I was investigating #6794 by running a local copy of ~finned-palmer, manually breaching by doing |pass [%j %ruin [~dinleb-rambep ~ ~]] and observing the ship hang for a very long time. When I interrupted the ship by doing ctrl+c, the problem became quite evident.

~finned-palmer was spending all its' time in +mo-filter-queue. Specifically, it was iterating over 450 000 blocked moves to an agent called %push-notify trying to find any moves from the ship that just breached. This is an agent from the long defunct %escape.

This PR enables us to recover from scenarios where userspace programs do crazy stuff like this. It does potentially break some invariants, but I believe this to be worth it because this is a manual intervention.

@belisarius222 @joemfb

@patosullivan
Copy link
Collaborator

Applied this fix to my live ship, seems to be working well.

@patosullivan
Copy link
Collaborator

Spoke too soon, the long hangs after breaches are still occurring w/this fix.

@pkova
Copy link
Collaborator Author

pkova commented Aug 28, 2024

You need to |nuke %push-notify to fix it after applying the code.

@patosullivan
Copy link
Collaborator

Alright, nuked it, will report back if I see the hanging again.

@patosullivan
Copy link
Collaborator

Ok, my ship just processed 15 breaches and it was only unresponsive for about ~4 minutes, normally that would be 15 * 15 minutes. I'm guessing it was just unresponsive due to the large number being processed back-to-back?

@patosullivan
Copy link
Collaborator

Could we also preemptively |nuke %push-notify so the effect ships don't all need to be told to do so?

Copy link
Collaborator

@joemfb joemfb left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this is good

@pkova pkova merged commit 425655a into develop Sep 4, 2024
@pkova pkova deleted the pkova/really-nuke branch September 4, 2024 16:09
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants