[Docs] explain cluster management + recovery operations

[joaofnfernandes]

@nicolaka commented on Tue Jan 31 2017

What's the problem/feature request?

just got off the phone with a customer who were very confused on node management with UCP. IMO both swarm and ucp docs lack the end-to-end story around cluster management. Neither docs goes through (in details) the difference between node rm node demote swarm leave and UCP's UI's node remove operations along with the recommended process to remove nodes from a cluster. For UCP especially, how does node operations affect the HA and what to do when node removal/demotion goes wrong.

I think fixing these issues are crucial for next release:

#1451
#815

What versions/components are affected?

2.1

---

@alexmavr commented on Tue Jan 31 2017

+1 I agree that our documentation is strongly lacking on that subject and I had a similar discussion earlier regarding several customer issues that arise from this confusion. Our recommended path for removing a node from the cluster / uninstalling UCP from one node should be:

1. Perform docker node demote for the target node from another manager node. This step can be skipped if the target node is not a manager
2. (Optional) wait for 5-10 seconds so that node reconciliation kicks in and starts cleaning up the UCP leftovers. This guarantees that adding new nodes later on is not going to result in problems.
3. Perform a docker swarm leave on the target node
4. Perform a docker node rm for the target node from another manager node.

---

@joaofnfernandes commented on Tue Jan 31

Relates to #2699

[alexmavr]

Fixed in #1792

@joaofnfernandes please close if you agree

[docker-robott]

Closed issues are locked after 30 days of inactivity.
This helps our team focus on active issues.

If you have found a problem that seems similar to this, please open a new issue.

/lifecycle locked