Runat contains bad value (monotonic clock) #2
Description
I've failed with usage of monotonic clock on linux within distributed system.
In general monotonic clock specified as "monotonic time since some unspecified starting point". On some system it equals to unix timestamp. But on other it contains in fact uptime of system. So, just by machine reboot (or having several replicated nodes) time consistency became broken
Clock will be changed to realtime clock.
I will consider timestamps, earlier than 10 years ago as monotonic timestamps
to avoid deletion of data with ttl tasks or early fireups of delayed tasks
(I don't believe in hosts with 10y uptime running tarantool with xqueue)
It is recommeded to upgrage all runat fields.
for updating space the following snipped may be used
local sp = box.space.spacename
local ix = sp.index.runat
local fn = ix.parts[1].fieldno
local clock = require'clock'
local log = require'log'
while true do
local c = 0
for _,t in ix:pairs({ 0 }, { iterator='GT' }):take(100) do
if t[fn] > 10*365*86400 then break end
local newval = t[fn] - clock.monotonic() + clock.realtime()
if newval < 10*365*86400 then error("Failed to correct runat") end
log.info("Fix %s time %s to %s",t[1],t[fn], newval)
sp:update({t[1]},{{'=',fn, newval }})
c=c+1
end
if c == 0 then break end;
end