Files
podman/libpod/container_graph.go
Matt Heon 46d874aa52 Refactor graph traversal & use for pod stop
First, refactor our existing graph traversal code to improve code
sharing. There still isn't much sharing between inward traversal
(stop, remove) and outward traversal (start) but stop and remove
are sharing most of their code, which seems a positive.

Second, add a new graph-traversal function to stop containers.
We already had start and remove; stop uses the newly-refactored
inward-traversal code which it shares with removal.

Third, rework the shared stop/removal inward-traversal code to
add locking. This allows parallel execution of stop and removal,
which should improve the performance of `podman pod rm` and
retain the performance of `podman pod stop` at about what it is
right now.

Fourth and finally, use the new graph-based stop when possible
to solve unordered stop problems with pods - specifically, the
infra container stopping before application containers, leaving
those containers without a working network.

Fixes https://issues.redhat.com/browse/RHEL-76827

Signed-off-by: Matt Heon <mheon@redhat.com>
2025-02-06 18:28:12 -05:00

529 lines
15 KiB
Go

//go:build !remote
package libpod
import (
"context"
"errors"
"fmt"
"strings"
"sync"
"github.com/containers/podman/v5/libpod/define"
"github.com/containers/podman/v5/pkg/parallel"
"github.com/sirupsen/logrus"
)
type containerNode struct {
lock sync.Mutex
id string
container *Container
dependsOn []*containerNode
dependedOn []*containerNode
}
// ContainerGraph is a dependency graph based on a set of containers.
type ContainerGraph struct {
nodes map[string]*containerNode
noDepNodes []*containerNode
notDependedOnNodes map[string]*containerNode
}
// DependencyMap returns the dependency graph as map with the key being a
// container and the value being the containers the key depends on.
func (cg *ContainerGraph) DependencyMap() (dependencies map[*Container][]*Container) {
dependencies = make(map[*Container][]*Container)
for _, node := range cg.nodes {
dependsOn := make([]*Container, len(node.dependsOn))
for i, d := range node.dependsOn {
dependsOn[i] = d.container
}
dependencies[node.container] = dependsOn
}
return dependencies
}
// BuildContainerGraph builds a dependency graph based on the container slice.
func BuildContainerGraph(ctrs []*Container) (*ContainerGraph, error) {
graph := new(ContainerGraph)
graph.nodes = make(map[string]*containerNode)
graph.notDependedOnNodes = make(map[string]*containerNode)
// Start by building all nodes, with no edges
for _, ctr := range ctrs {
ctrNode := new(containerNode)
ctrNode.id = ctr.ID()
ctrNode.container = ctr
graph.nodes[ctr.ID()] = ctrNode
graph.notDependedOnNodes[ctr.ID()] = ctrNode
}
// Now add edges based on dependencies
for _, node := range graph.nodes {
deps := node.container.Dependencies()
for _, dep := range deps {
// Get the dep's node
depNode, ok := graph.nodes[dep]
if !ok {
return nil, fmt.Errorf("container %s depends on container %s not found in input list: %w", node.id, dep, define.ErrNoSuchCtr)
}
// Add the dependent node to the node's dependencies
// And add the node to the dependent node's dependedOn
node.dependsOn = append(node.dependsOn, depNode)
depNode.dependedOn = append(depNode.dependedOn, node)
// The dependency now has something depending on it
delete(graph.notDependedOnNodes, dep)
}
// Maintain a list of nodes with no dependencies
// (no edges coming from them)
if len(deps) == 0 {
graph.noDepNodes = append(graph.noDepNodes, node)
}
}
// Need to do cycle detection
// We cannot start or stop if there are cyclic dependencies
cycle, err := detectCycles(graph)
if err != nil {
return nil, err
} else if cycle {
return nil, fmt.Errorf("cycle found in container dependency graph: %w", define.ErrInternal)
}
return graph, nil
}
// Detect cycles in a container graph using Tarjan's strongly connected
// components algorithm
// Return true if a cycle is found, false otherwise
func detectCycles(graph *ContainerGraph) (bool, error) {
type nodeInfo struct {
index int
lowLink int
onStack bool
}
index := 0
nodes := make(map[string]*nodeInfo)
stack := make([]*containerNode, 0, len(graph.nodes))
var strongConnect func(*containerNode) (bool, error)
strongConnect = func(node *containerNode) (bool, error) {
logrus.Debugf("Strongconnecting node %s", node.id)
info := new(nodeInfo)
info.index = index
info.lowLink = index
index++
nodes[node.id] = info
stack = append(stack, node)
info.onStack = true
logrus.Debugf("Pushed %s onto stack", node.id)
// Work through all nodes we point to
for _, successor := range node.dependsOn {
if _, ok := nodes[successor.id]; !ok {
logrus.Debugf("Recursing to successor node %s", successor.id)
cycle, err := strongConnect(successor)
if err != nil {
return false, err
} else if cycle {
return true, nil
}
successorInfo := nodes[successor.id]
if successorInfo.lowLink < info.lowLink {
info.lowLink = successorInfo.lowLink
}
} else {
successorInfo := nodes[successor.id]
if successorInfo.index < info.lowLink && successorInfo.onStack {
info.lowLink = successorInfo.index
}
}
}
if info.lowLink == info.index {
l := len(stack)
if l == 0 {
return false, fmt.Errorf("empty stack in detectCycles: %w", define.ErrInternal)
}
// Pop off the stack
topOfStack := stack[l-1]
stack = stack[:l-1]
// Popped item is no longer on the stack, mark as such
topInfo, ok := nodes[topOfStack.id]
if !ok {
return false, fmt.Errorf("finding node info for %s: %w", topOfStack.id, define.ErrInternal)
}
topInfo.onStack = false
logrus.Debugf("Finishing node %s. Popped %s off stack", node.id, topOfStack.id)
// If the top of the stack is not us, we have found a
// cycle
if topOfStack.id != node.id {
return true, nil
}
}
return false, nil
}
for id, node := range graph.nodes {
if _, ok := nodes[id]; !ok {
cycle, err := strongConnect(node)
if err != nil {
return false, err
} else if cycle {
return true, nil
}
}
}
return false, nil
}
// Visit a node on a container graph and start the container, or set an error if
// a dependency failed to start. if restart is true, startNode will restart the node instead of starting it.
func startNode(ctx context.Context, node *containerNode, setError bool, ctrErrors map[string]error, ctrsVisited map[string]bool, restart bool) {
// First, check if we have already visited the node
if ctrsVisited[node.id] {
return
}
// If setError is true, a dependency of us failed
// Mark us as failed and recurse
if setError {
// Mark us as visited, and set an error
ctrsVisited[node.id] = true
ctrErrors[node.id] = fmt.Errorf("a dependency of container %s failed to start: %w", node.id, define.ErrCtrStateInvalid)
// Hit anyone who depends on us, and set errors on them too
for _, successor := range node.dependedOn {
startNode(ctx, successor, true, ctrErrors, ctrsVisited, restart)
}
return
}
// Have all our dependencies started?
// If not, don't visit the node yet
depsVisited := true
for _, dep := range node.dependsOn {
depsVisited = depsVisited && ctrsVisited[dep.id]
}
if !depsVisited {
// Don't visit us yet, all dependencies are not up
// We'll hit the dependencies eventually, and when we do it will
// recurse here
return
}
// Going to try to start the container, mark us as visited
ctrsVisited[node.id] = true
ctrErrored := false
// Check if dependencies are running
// Graph traversal means we should have started them
// But they could have died before we got here
// Does not require that the container be locked, we only need to lock
// the dependencies
depsStopped, err := node.container.checkDependenciesRunning()
if err != nil {
ctrErrors[node.id] = err
ctrErrored = true
} else if len(depsStopped) > 0 {
// Our dependencies are not running
depsList := strings.Join(depsStopped, ",")
ctrErrors[node.id] = fmt.Errorf("the following dependencies of container %s are not running: %s: %w", node.id, depsList, define.ErrCtrStateInvalid)
ctrErrored = true
}
// Lock before we start
node.container.lock.Lock()
// Sync the container to pick up current state
if !ctrErrored {
if err := node.container.syncContainer(); err != nil {
ctrErrored = true
ctrErrors[node.id] = err
}
}
// Start the container (only if it is not running)
if !ctrErrored && len(node.container.config.InitContainerType) < 1 {
if !restart && node.container.state.State != define.ContainerStateRunning {
if err := node.container.initAndStart(ctx); err != nil {
ctrErrored = true
ctrErrors[node.id] = err
}
}
if restart && node.container.state.State != define.ContainerStatePaused && node.container.state.State != define.ContainerStateUnknown {
if err := node.container.restartWithTimeout(ctx, node.container.config.StopTimeout); err != nil {
ctrErrored = true
ctrErrors[node.id] = err
}
}
}
node.container.lock.Unlock()
// Recurse to anyone who depends on us and start them
for _, successor := range node.dependedOn {
startNode(ctx, successor, ctrErrored, ctrErrors, ctrsVisited, restart)
}
}
// Contains all details required for traversing the container graph.
type nodeTraversal struct {
// Protects reads and writes to the two maps.
lock sync.Mutex
// Optional. but *MUST* be locked.
// Should NOT be changed once a traversal is started.
pod *Pod
// Function to execute on the individual container being acted on.
// Should NOT be changed once a traversal is started.
actionFunc func(ctr *Container, pod *Pod) error
// Shared list of errors for all containers currently acted on.
ctrErrors map[string]error
// Shared list of what containers have been visited.
ctrsVisited map[string]bool
}
// Perform a traversal of the graph in an inwards direction - meaning from nodes
// with no dependencies, recursing inwards to the nodes they depend on.
// Safe to run in parallel on multiple nodes.
func traverseNodeInwards(node *containerNode, nodeDetails *nodeTraversal, setError bool) {
node.lock.Lock()
// If we already visited this node, we're done.
nodeDetails.lock.Lock()
visited := nodeDetails.ctrsVisited[node.id]
nodeDetails.lock.Unlock()
if visited {
node.lock.Unlock()
return
}
// Someone who depends on us failed.
// Mark us as failed and recurse.
if setError {
nodeDetails.lock.Lock()
nodeDetails.ctrsVisited[node.id] = true
nodeDetails.ctrErrors[node.id] = fmt.Errorf("a container that depends on container %s could not be stopped: %w", node.id, define.ErrCtrStateInvalid)
nodeDetails.lock.Unlock()
node.lock.Unlock()
// Hit anyone who depends on us, set errors there as well.
for _, successor := range node.dependsOn {
traverseNodeInwards(successor, nodeDetails, true)
}
return
}
// Does anyone still depend on us?
// Cannot stop if true. Once all our dependencies have been stopped,
// we will be stopped.
for _, dep := range node.dependedOn {
// The container that depends on us hasn't been removed yet.
// OK to continue on
nodeDetails.lock.Lock()
ok := nodeDetails.ctrsVisited[dep.id]
nodeDetails.lock.Unlock()
if !ok {
node.lock.Unlock()
return
}
}
ctrErrored := false
if err := nodeDetails.actionFunc(node.container, nodeDetails.pod); err != nil {
ctrErrored = true
nodeDetails.lock.Lock()
nodeDetails.ctrErrors[node.id] = err
nodeDetails.lock.Unlock()
}
// Mark as visited *only after* finished with operation.
// This ensures that the operation has completed, one way or the other.
// If an error was set, only do this after the viral ctrErrored
// propagates in traverseNodeInwards below.
// Same with the node lock - we don't want to release it until we are
// marked as visited.
if !ctrErrored {
nodeDetails.lock.Lock()
nodeDetails.ctrsVisited[node.id] = true
nodeDetails.lock.Unlock()
node.lock.Unlock()
}
// Recurse to anyone who we depend on and work on them
for _, successor := range node.dependsOn {
traverseNodeInwards(successor, nodeDetails, ctrErrored)
}
// If we propagated an error, finally mark us as visited here, after
// all nodes we traverse to have already been marked failed.
// If we don't do this, there is a race condition where a node could try
// and perform its operation before it was marked failed by the
// traverseNodeInwards triggered by this process.
if ctrErrored {
nodeDetails.lock.Lock()
nodeDetails.ctrsVisited[node.id] = true
nodeDetails.lock.Unlock()
node.lock.Unlock()
}
}
// Stop all containers in the given graph, assumed to be a graph of pod.
// Pod is mandatory and should be locked.
func stopContainerGraph(ctx context.Context, graph *ContainerGraph, pod *Pod, timeout *uint, cleanup bool) (map[string]error, error) {
// Are there actually any containers in the graph?
// If not, return immediately.
if len(graph.nodes) == 0 {
return map[string]error{}, nil
}
nodeDetails := new(nodeTraversal)
nodeDetails.pod = pod
nodeDetails.ctrErrors = make(map[string]error)
nodeDetails.ctrsVisited = make(map[string]bool)
traversalFunc := func(ctr *Container, pod *Pod) error {
ctr.lock.Lock()
defer ctr.lock.Unlock()
if err := ctr.syncContainer(); err != nil {
return err
}
realTimeout := ctr.config.StopTimeout
if timeout != nil {
realTimeout = *timeout
}
if err := ctr.stop(realTimeout); err != nil && !errors.Is(err, define.ErrCtrStateInvalid) && !errors.Is(err, define.ErrCtrStopped) {
return err
}
if cleanup {
return ctr.fullCleanup(ctx, false)
}
return nil
}
nodeDetails.actionFunc = traversalFunc
doneChans := make([]<-chan error, 0, len(graph.notDependedOnNodes))
// Parallel enqueue jobs for all our starting nodes.
if len(graph.notDependedOnNodes) == 0 {
return nil, fmt.Errorf("no containers in pod %s are not dependencies of other containers, unable to stop", pod.ID())
}
for _, node := range graph.notDependedOnNodes {
doneChan := parallel.Enqueue(ctx, func() error {
traverseNodeInwards(node, nodeDetails, false)
return nil
})
doneChans = append(doneChans, doneChan)
}
// We don't care about the returns values, these functions always return nil
// But we do need all of the parallel jobs to terminate.
for _, doneChan := range doneChans {
<-doneChan
}
return nodeDetails.ctrErrors, nil
}
// Remove all containers in the given graph
// Pod is optional, and must be locked if given.
func removeContainerGraph(ctx context.Context, graph *ContainerGraph, pod *Pod, timeout *uint, force bool) (map[string]*ContainerNamedVolume, map[string]bool, map[string]error, error) {
// Are there actually any containers in the graph?
// If not, return immediately.
if len(graph.nodes) == 0 {
return nil, nil, nil, nil
}
nodeDetails := new(nodeTraversal)
nodeDetails.pod = pod
nodeDetails.ctrErrors = make(map[string]error)
nodeDetails.ctrsVisited = make(map[string]bool)
ctrNamedVolumes := make(map[string]*ContainerNamedVolume)
traversalFunc := func(ctr *Container, pod *Pod) error {
ctr.lock.Lock()
defer ctr.lock.Unlock()
if err := ctr.syncContainer(); err != nil {
return err
}
for _, vol := range ctr.config.NamedVolumes {
ctrNamedVolumes[vol.Name] = vol
}
if pod != nil && pod.state.InfraContainerID == ctr.ID() {
pod.state.InfraContainerID = ""
if err := pod.save(); err != nil {
return fmt.Errorf("error removing infra container %s from pod %s: %w", ctr.ID(), pod.ID(), err)
}
}
opts := ctrRmOpts{
Force: force,
RemovePod: true,
Timeout: timeout,
}
if _, _, err := ctr.runtime.removeContainer(ctx, ctr, opts); err != nil {
return err
}
return nil
}
nodeDetails.actionFunc = traversalFunc
doneChans := make([]<-chan error, 0, len(graph.notDependedOnNodes))
// Parallel enqueue jobs for all our starting nodes.
if len(graph.notDependedOnNodes) == 0 {
return nil, nil, nil, fmt.Errorf("no containers in graph are not dependencies of other containers, unable to stop")
}
for _, node := range graph.notDependedOnNodes {
doneChan := parallel.Enqueue(ctx, func() error {
traverseNodeInwards(node, nodeDetails, false)
return nil
})
doneChans = append(doneChans, doneChan)
}
// We don't care about the returns values, these functions always return nil
// But we do need all of the parallel jobs to terminate.
for _, doneChan := range doneChans {
<-doneChan
}
return ctrNamedVolumes, nodeDetails.ctrsVisited, nodeDetails.ctrErrors, nil
}