Skip to content
This repository has been archived by the owner on Jan 30, 2020. It is now read-only.

engine role can float around constantly #639

Closed
bcwaldon opened this issue Jul 10, 2014 · 4 comments · Fixed by #640
Closed

engine role can float around constantly #639

bcwaldon opened this issue Jul 10, 2014 · 4 comments · Fixed by #640
Assignees
@bcwaldon
Labels
kind/cleanup priority/P1
Milestone
v0.5.4

Comments

@bcwaldon
Copy link
Contributor

Every fleet daemon will attempt to acquire a lock in etcd every 2s. If a given daemon can acquire this lock, it will run through a single engine reconciliation loop. This has two flaws:

  1. the role can move from machine to machine, making it difficult to debug
  2. the cluster can effectively reconcile constantly

We should basically treat the engine role as a lease. The lease should last much longer than 2s (maybe 30s), and the engine should simply renew it.
@bcwaldon bcwaldon changed the title engine role can around constantly engine role can float around constantly Jul 10, 2014
@bcwaldon bcwaldon self-assigned this Jul 10, 2014
@jonboulle
Copy link
Contributor

engine role can float around

float

@bcwaldon
Copy link
Contributor Author

@jonboulle That gif sums up this issue perfectly.

@carmstrong
Copy link

That engine looks seized up

@bcwaldon bcwaldon mentioned this issue Jul 10, 2014
Merged
@bcwaldon
Copy link
Contributor Author

@carmstrong I'm workin on it! #640

@bcwaldon bcwaldon added this to the v0.5.4 milestone Jul 14, 2014
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Assignees

@bcwaldon bcwaldon

Labels
kind/cleanup priority/P1
Projects
None yet
Milestone
v0.5.4
Development

Successfully merging a pull request may close this issue.

Switch from engine "lock" to "leased" engine leader role bcwaldon/fleet
3 participants
@carmstrong @bcwaldon @jonboulle