| Commit message (Collapse) | Author | Age | Files | Lines |
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
| |
While we wait for a peer's HELLO_REPLY, anything else that comes from it
makes our HELLO go again at once, at most once a second, instead of when
the backoff says.
|
| |
|
|
|
|
| |
When the answer to a pull is overdue (the sync reply timeout: 1 s in the
simtests, 10 s in production), the pull gives up the slot, and the next
waiting neighbour is pulled at once.
|
| |
|
|
|
|
|
| |
The heal clock now also restarts when our heads and the peer's have both
changed since the last STATUS that showed them differing. While writes
go on, two neighbours's heads differ because of changes still on their
way, and that alone no longer leads to a heal.
|
| |
|
|
|
|
|
|
| |
Before pulling, a node waits one measured round trip to the announcer
(1 s at most) and pulls only what is still missing. If the round trip
isn't known yet, it pulls at once, as now. On the clean benchmark links
the round trip is a few milliseconds, so those scenarios shouldn't
change.
|
| |
|
|
|
|
|
| |
A neighbour now answers an announcement only if it lacks some of the
announced heads. Heads that merely differ are no longer enough, since
they do while writes go on. The check reads the document, so it runs
off the message consumer, as the pull check already does.
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
|