Files
moby/vendor/github.com/docker/swarmkit/manager/state/doc.go
Yong Tang ec860ca252 Update SwarmKit to 6bc357e9c5f0ac2cdf801898a43d08c260b4d5d0
This fix tries to update the SwarmKit from
ed384f3b3957f65e3111bd020f9815f3d4296fa2
to
6bc357e9c5f0ac2cdf801898a43d08c260b4d5d0

The following is the list of docker related changes:
1. Took long time for Docker Swarm service turn desired state from Ready to Running (Issue #28291)
2. Native Swarm in 1.12 - panic: runtime error: index out of range (Issue #25608)
3. Global mode target replicas keep increasing (Issue #30854)
4. Creating service with publish mode=host and without published port crashes swarm manager (Issue #30938)
5. Define signals used to stop containers for updates (Issue #25696) (PR #30754)

This fix fixes #28291, #25608, #30854, #30938.
This fix is required by PR #30754.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
2017-02-16 08:50:00 -08:00

33 lines
1.4 KiB
Go

// Package state provides interfaces to work with swarm cluster state.
//
// The primary interface is Store, which abstracts storage of this cluster
// state. Store exposes a transactional interface for both reads and writes.
// To perform a read transaction, View accepts a callback function that it
// will invoke with a ReadTx object that gives it a consistent view of the
// state. Similarly, Update accepts a callback function that it will invoke with
// a Tx object that allows reads and writes to happen without interference from
// other transactions.
//
// This is an example of making an update to a Store:
//
// err := store.Update(func(tx state.Tx) {
// if err := tx.Nodes().Update(newNode); err != nil {
// return err
// }
// return nil
// })
// if err != nil {
// return fmt.Errorf("transaction failed: %v", err)
// }
//
// WatchableStore is a version of Store that exposes watch functionality.
// These expose a publish/subscribe queue where code can subscribe to
// changes of interest. This can be combined with the ViewAndWatch function to
// "fork" a store, by making a snapshot and then applying future changes
// to keep the copy in sync. This approach lets consumers of the data
// use their own data structures and implement their own concurrency
// strategies. It can lead to more efficient code because data consumers
// don't necessarily have to lock the main data store if they are
// maintaining their own copies of the state.
package state