DyrellC
commented
5 years ago @jakubcech @GalRogozinski
I have found a way to trigger the resetting corrupted milestone service consistently. While spamming milestones at my node using my spammer (not compass) I was able to trigger the mechanism by interrupting one of the attachToTangle() instances for a milestone. I spammed 60 milestones in total using my spammer tools and interrupted the 21st one from attaching correctly. By only interrupting the one milestone the tangle was no longer able to become solid beyond milestone 20, but the other 40 milestones were able to be introduced to the db with a bunch of zero value transactions as well. Once I attached a 21st milestone, the resetting mechanism kicked in and managed to reset the remaining milestones and solidify.
This was in a zero value transaction environment, my next test will be to see what happens when there are transactions with value present in both the first 20 milestones prior to the failed milestone, and the remaining transactions attached after the failed milestone to see what happens in this instance. I expect that what will happen is that if there are no conflicting spends then all the milestones will solidify again, but if I introduce conflicts, it should invalidate the milestones that tried to confirm the inconsistent transactions following the missed milestone, and will instead solidify to milestone 21 and no further.
GalRogozinski
rajivshah3
jakubcech
Description
Investigate when the
repairing corrupted milestonemechanism triggers nowadays. We know (sorta) it keeps triggering still, but we're not sure if it works when it does. And we especially aren't sure what the exact scenario of when this happens is exactly. One reason might be that a newer milestone solidifies faster than an older one.This could be related to #1434.