What happens when a machine handling a shard goes down?


#1

Design Cache Q: What happens when a machine handling a shard goes down?


#2

Those cached values are lost unless we have a replica and restore from the replica..


#3

what does it mean by storing change log with version number?


#4

master write to change logs and sent over to slaves to sync up. so even slave(s) are down for a period of time, slaves can still caught up with all missed updates in change logs. version number with change logs so slaves know which ones were already applied and which ones are not applied yet.


#5

"If we are fine with all servers becoming eventually consistent, then we can have one master ( taking all the write traffic ) and many slaves where slaves can service the read traffic as well."


#6

What write traffic? The only traffic we have is read traffic. write is an internal process in case of a miss.


#7

Write traffic here means adding new data(or updating the existing) to the master.