5.1.b-Leader Election--Fundamentals--Leader election via a lease

This video shows how to implement leader election using a lease mechanism.

[@mostinho7]

Done thanks We don’t want to implement leader election ourselves due to race condition on crashes, so we use one of these mechanisms. 0:10 Lease (all instanced try to be the leader): 1:47 a new task shows up in the db with a work time and status (done or not done) Each task has “leases” for the leader election

[@user-eg7yz4ub6u]

I wonder what magic makes leasee attribute to disappear when service 1 crushed? In my experience some timeout attributes usually are used for other services to understand that the lease is expired

[@JeffreyRichter]

Yes, the lease expires due to timeout

[@TheHellst0rm]

i wonder how does the work item appear in the database? Who puts it there? Doesnt it cause the same problem with leadership again?

[@JeffreyRichter]

Usually a customer initiates some workflow which sends a request to a service which passes through a load balancer in order to select 1 server which then goes and adds an entry to a database. So, in this scenario, the load balancer effectively performs the leader election.