As we talk on dev group and some tickets here and in skywire repo, we decided to do two thing, re-implement monitors and add heartbeat to skywire for services as bellow:
reimplement dmsg-monitor and ar-monitor just based on UT entries. Because of 5 minute delay on UT update will use temp-dead-entries logic, which we will have double check in 10 minute and if both check the entry was dead then remove it.
reimplement tpd-monitor based on dmsg entries and ar entries. We make sure dmsg-monitor and ar-monitor work well.
sd-monitor first remove entries that not online on UT, by temp-dead-entries scenario, and for available entries, we should trying to connect same as before.
Also we should implement heartbeat logic for dsmg, ar and all sd entries on visor that send request as re-register to services each 10 minute (or something else, we should pass it as flag to visor).
As we talk on dev group and some tickets here and in skywire repo, we decided to do two thing, re-implement monitors and add heartbeat to skywire for services as bellow:
temp-dead-entrieslogic, which we will have double check in 10 minute and if both check the entry was dead then remove it.temp-dead-entriesscenario, and for available entries, we should trying to connect same as before.Also we should implement heartbeat logic for dsmg, ar and all sd entries on visor that send request as re-register to services each 10 minute (or something else, we should pass it as flag to visor).