re-add dead hosts when all hosts are unavailable
Bug #279883 reported by
Jehiah
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
txLoadBalancer |
Fix Committed
|
High
|
Duncan McGreggor |
Bug Description
When a service has a set of hosts (or only one) and any hosts have connection failures and are marked as such so they don't get new incoming connections, they will normally be checked at a regular interval by hostchecker to be re-added to the available pool of hosts.
However, if *all* the hosts for a group are removed from rotation due to failures, there should be additional handling so that new connections re-try failed hosts instead of refusing connections until hostchecker's next run (if it's even enabled).
we need to do is add some logic in the host-checking code: in the
even that all hosts are marked as dead, mark all as active.
Changed in txloadbalancer: | |
assignee: | nobody → oubiwann |
importance: | Undecided → High |
To post a comment you must log in.
Jehiah,
I've added code that does this now. It was committed in trunk, r209.