Abnormally high delta on single reflector #82
Replies: 5 comments 55 replies
-
Ok, I have some ideas here. First off, because ICMP timestamps are "milliseconds since midnight UTC" it's inevitable we'll have to deal with rollover issues... This is annoying but something we haven't tackled yet so it's not surprising we have a problem. Of course, if only one rolls over at a time, the voting by taking the 3rd of 5 should fix it... but if 3 roll over we'll have persistent ridiculous delays that take a while to decay. Several thoughts:
|
Beta Was this translation helpful? Give feedback.
-
Ok, I pushed a proposed scaffolding for the error handling to this branch... it's just got comments where we'd need to trigger reselection, and it's got a spot where it'd like to "goto" a label and skip over what we'd normally do if we had good data... but I don't know if we can rely on goto since it's 5.2 take a look and see what you think, we can revert or modify as needed |
Beta Was this translation helpful? Give feedback.
-
Ok back from my thing, @Fail-Safe does the current version run for you without problems in the "normal" cases at least (we will have to come up with a way to test midnight rollovers separately). |
Beta Was this translation helpful? Give feedback.
-
@dlakelan FYI, seeing this (see 'Reselect signaled') under "normal" usage now:
|
Beta Was this translation helpful? Give feedback.
-
Per @Lochnair:
Beta Was this translation helpful? Give feedback.
All reactions