Set pkt retry to 2 (was 1) for hub nodes. (#1125)

Too many retries appears to cause network congestion issues slowing down
links with multiple hops, but reducing this to 1 then caused single links
to have poor performance. Try a rety of 2 as a better compromise.
This commit is contained in:
Tim Wilkinson 2024-03-25 21:08:30 -07:00 committed by GitHub
parent 8b6339a099
commit 2fc76191a0
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 7 additions and 6 deletions

View File

@ -50,9 +50,10 @@ local dtd_distance = 50 -- distance (meters) after which nodes connected with Dt
local connect_timeout = 5 -- timeout (seconds) when fetching information from other nodes
local speed_time = 10 --
local speed_limit = 1000 -- close connection if it's too slow (< 1kB/s for 10 seconds)
local default_retries_scale = 2
local default_short_retries = 6
local default_long_retries = 4
local default_short_retries = 6 -- Default retries for a node with a single connection
local default_long_retries = 4 --
local minimum_short_retries = 2 -- Minimum retries for nodes with multiple connections
local minimum_long_retries = 2 --
local NFT = "/usr/sbin/nft"
local IW = "/usr/sbin/iw"
@ -966,13 +967,13 @@ function lqm()
hidden_nodes = hidden
-- Adjust retry attempts. If we're just managing a single connection, we can retry failed transmissions.
-- If we're managing many, we can't afford the delays associated with retries, to reduce them to the minimum.
-- If we're managing many, we can't afford the delays associated with retries, so reduce them to the minimum.
-- Don't retry when distance is max because the delay makes thing unusable and blocks other closer traffic.
local short_retries
local long_retries
if distance == config.max_distance or rfcount > 1 then
short_retries = 1
long_retries = 1
short_retries = minimum_short_retries
long_retries = minimum_long_retries
else
short_retries = default_short_retries
long_retries = default_long_retries