Simply the watchdog by fixing various timeouts. (#1038)

Unfortunately there doesnt appear to be much flexibility in the various
hardware watchdogs on radios, so setting the watchdog > 60 seconds mostly
doesnt work. So rework the settings to allow for this and that our watchdog
tests must be frequent and quick.
This commit is contained in:
Tim Wilkinson 2023-12-20 00:25:12 -08:00 committed by GitHub
parent 2fe3904f0a
commit e5bd9bc066
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 26 additions and 48 deletions

View File

@ -54,12 +54,12 @@ function wait_for_ticks(ticks)
local when = nixio.sysinfo().uptime + ticks
while true
do
ticks = when - nixio.sysinfo().uptime
if ticks > 0 then
if ticks >= 0 then
coroutine.yield(ticks)
else
break
end
ticks = when - nixio.sysinfo().uptime
end
end

View File

@ -39,15 +39,9 @@ local REBOOT = "/sbin/reboot"
local W = {}
-- Configuration limits and defaults
local config_limits = {
startup_delay = { 600, 600, 3600 },
ping_count = { 1, 3, 10 },
ping_timeout = { 1, 5, 10 },
tick = { 60, 120, 600 },
failures = { 2, 5, 25 },
daily = { -1, -1, 23 }
}
local tick = 20
local ping_timeout = 3
local startup_delay = 600
-- Set of daemons to monitor
local default_daemons = "olsrd dnsmasq telnetd dropbear uhttpd"
@ -68,6 +62,7 @@ function W.get_config(verbose)
ping_addresses[#ping_addresses + 1] = address
end
end
local daemons = {}
local mydaemons = c:get("aredn", "@watchdog[0]", "daemons") or default_daemons
for daemon in mydaemons:gmatch("(%S+)") do
@ -76,34 +71,14 @@ function W.get_config(verbose)
end
daemons[#daemons + 1] = daemon
end
local config = {
local daily = tonumber(c:get("aredn", "@watchdog[0]", "daily") or nil) or -1
return {
ping_addresses = ping_addresses,
daemons = daemons
daemons = daemons,
daily = daily
}
for k, v in pairs(config_limits)
do
local val = tonumber(c:get("aredn", "@watchdog[0]", k) or nil)
if not val then
config[k] = v[2]
elseif val < v[1] then
config[k] = v[1]
elseif val > v[3] then
config[k] = v[3]
else
config[k] = val
end
end
-- Make sure we have enough tick time for any pings
local total_ping_time = 30 + (config.ping_timeout + config.ping_count) * #config.ping_addresses
if total_ping_time > config.tick then
config.tick = math.ceil(total_ping_time / 60) * 60
if verbose then
mainlog:write("adjusting tick to " .. config.tick)
end
end
return config
end
function W.start()
@ -112,12 +87,15 @@ function W.start()
return
end
local ub = ubus.connect()
local config = W.get_config(true)
-- Dont start monitoring too soon. Let the system settle down.
wait_for_ticks(math.max(1, config.startup_delay - nixio.sysinfo().uptime))
ub:call("system", "watchdog", { frequency = 1 })
ub:call("system", "watchdog", { timeout = 60 })
-- Dont start monitoring too soon. Let the system settle down.
wait_for_ticks(math.max(0, startup_delay - nixio.sysinfo().uptime))
local ub = ubus.connect()
ub:call("system", "watchdog", { magicclose = true })
ub:call("system", "watchdog", { stop = true })
@ -129,9 +107,6 @@ function W.start()
return
end
-- We make sure it's at least 5 minutes
ub:call("system", "watchdog", { timeout = math.ceil(math.max(300, config.tick * config.failures)) })
local daily_reboot_armed = false
while true
@ -146,7 +121,7 @@ function W.start()
-- over we must have just seen the previous hour
if config.daily ~= -1 then
local time = os.date("*t")
if time.min >= (60 - config.tick * 3) and (time.hour + 1) % 24 == config.daily then
if time.min >= 55 and (time.hour + 1) % 24 == config.daily then
daily_reboot_armed = true
elseif daily_reboot_armed and time.hour == config.daily then
mainlog:write("reboot")
@ -177,7 +152,7 @@ function W.start()
success = false
for _, address in ipairs(config.ping_addresses)
do
if os.execute(PING .. " -c " .. config.ping_count .. " -A -q -W " .. config.ping_timeout .. " " .. address .. " > /dev/null 2>&1") == 0 then
if os.execute(PING .. " -c 1 -A -q -W " .. ping_timeout .. " " .. address .. " > /dev/null 2>&1") == 0 then
success = true
break
else
@ -191,12 +166,13 @@ function W.start()
end
if success then
wd:write("V")
wd:write("1")
wd:flush()
else
mainlog:write("failed")
end
wait_for_ticks(math.max(1, config.tick - (os.time() - now)))
wait_for_ticks(math.max(0, tick - (os.time() - now)))
end
end

View File

@ -228,7 +228,7 @@ function lqm()
end
-- Let things startup for a while before we begin
wait_for_ticks(math.max(1, 30 - nixio.sysinfo().uptime))
wait_for_ticks(math.max(0, 30 - nixio.sysinfo().uptime))
-- Create filters (cannot create during install as they disappear on reboot)
os.execute(NFT .. " flush chain ip fw4 input_lqm 2> /dev/null")
@ -599,6 +599,7 @@ function lqm()
local raw = io.popen("/usr/bin/curl --retry 0 --connect-timeout " .. connect_timeout .. " --speed-time " .. speed_time .. " --speed-limit " .. speed_limit .. " -s \"http://" .. track.ip .. ":8080/cgi-bin/sysinfo.json?link_info=1&lqm=1\" -o - 2> /dev/null")
local info = luci.jsonc.parse(raw:read("*a"))
raw:close()
wait_for_ticks(0)
if info then
rflinks[track.mac] = nil
if tonumber(info.lat) and tonumber(info.lon) then
@ -716,6 +717,7 @@ function lqm()
end
ptime = socket.gettime(0) - pstart
sigsock:close()
wait_for_ticks(0)
local ping_loss_run_avg = 1 - config.ping_penalty / 100
if success > 0 then