improve sync latency significantly by using goroutines

This commit is contained in:
2026-06-01 21:17:51 +00:00
parent a98745dd17
commit f1deacee10
3 changed files with 204 additions and 202 deletions
+199 -199
View File
@@ -5,8 +5,11 @@ import (
"fmt"
"log"
"strings"
"time"
paas "proxmoxaas-common-lib"
"golang.org/x/sync/errgroup"
)
func (cluster *Cluster) Init(pve ProxmoxClient) {
@@ -14,287 +17,284 @@ func (cluster *Cluster) Init(pve ProxmoxClient) {
}
func (cluster *Cluster) Get() (*Cluster, error) {
cluster_ch := make(chan *Cluster)
err_ch := make(chan error)
// aquire cluster lock
cluster.lock.Lock()
defer cluster.lock.Unlock()
go func() {
// aquire cluster lock
cluster.lock.Lock()
defer cluster.lock.Unlock()
cluster_ch <- cluster
err_ch <- nil
}()
return <-cluster_ch, <-err_ch
return cluster, nil
}
// hard sync cluster
func (cluster *Cluster) Sync() error {
err_ch := make(chan error)
// aquire lock on cluster, release on return
cluster.lock.Lock()
go func() {
// aquire lock on cluster, release on return
cluster.lock.Lock()
cluster.Nodes = make(map[string]*Node)
cluster.Nodes = make(map[string]*Node)
wg, _ := errgroup.WithContext(context.Background())
// get all nodes
nodes, err := cluster.pve.Nodes()
if err != nil {
err_ch <- err
return
}
// for each node:
for _, hostName := range nodes {
// get all nodes
nodes, err := cluster.pve.Nodes()
if err != nil {
return err
}
// for each node:
for _, hostName := range nodes {
wg.Go(func() error {
start := time.Now()
// rebuild node
err := cluster.RebuildNode(hostName)
if err != nil { // if an error was encountered, continue and log the error
log.Printf("[ERR ] %s", err)
} else { // otherwise log success
log.Printf("[INFO] successfully synced node %s", hostName)
log.Printf("[ERR ] error encountered while syncing node %s: %s", hostName, err)
} else {
log.Printf("[INFO] synced node %s in %d ms", hostName, time.Since(start).Milliseconds())
}
}
return err
})
}
cluster.lock.Unlock()
err = wg.Wait()
if err != nil {
return err
}
err = cluster.ResolvePoolMembership()
if err != nil {
err_ch <- err
}
cluster.lock.Unlock()
err_ch <- nil
}()
err = cluster.ResolvePoolMembership()
if err != nil {
return err
}
return <-err_ch
return nil
}
func (cluster *Cluster) ResolvePoolMembership() error {
err_ch := make(chan error)
// aquire lock on cluster, release on return
cluster.lock.Lock()
go func() {
// aquire lock on cluster, release on return
cluster.lock.Lock()
//resolve pool membership
pools, err := cluster.pve.client.Pools(context.Background())
//resolve pool membership
pools, err := cluster.pve.client.Pools(context.Background())
if err != nil {
return err
}
for _, pool := range pools {
pool, err = cluster.pve.client.Pool(context.Background(), pool.PoolID)
if err != nil {
err_ch <- err
return err
}
for _, pool := range pools {
pool, err = cluster.pve.client.Pool(context.Background(), pool.PoolID)
if err != nil {
err_ch <- err
}
for _, member := range pool.Members {
if member.Type == "lxc" || member.Type == "qemu" {
node, ok := cluster.Nodes[member.Node]
if !ok {
err_ch <- fmt.Errorf("Instance %d has no node", member.VMID)
}
instance, ok := node.Instances[InstanceID(member.VMID)]
if !ok {
err_ch <- fmt.Errorf("Instance %d claimed to be in node %s but was not", member.VMID, node.Name)
}
instance.Pool = pool.PoolID
log.Printf("[INFO] successfully resolved pool membership for vmid=%d pool=%s", member.VMID, pool.PoolID)
for _, member := range pool.Members {
if member.Type == "lxc" || member.Type == "qemu" {
node, ok := cluster.Nodes[member.Node]
if !ok {
return fmt.Errorf("Instance %d has no node", member.VMID)
}
instance, ok := node.Instances[InstanceID(member.VMID)]
if !ok {
return fmt.Errorf("Instance %d claimed to be in node %s but was not", member.VMID, node.Name)
}
instance.Pool = pool.PoolID
log.Printf("[INFO] resolved pool membership for vmid=%d pool=%s", member.VMID, pool.PoolID)
}
}
}
err_ch <- nil
cluster.lock.Unlock()
}()
return <-err_ch
cluster.lock.Unlock()
return nil
}
// get a node in the cluster
func (cluster *Cluster) GetNode(hostName string) (*Node, error) {
host_ch := make(chan *Node)
err_ch := make(chan error)
// aquire cluster lock
cluster.lock.Lock()
defer cluster.lock.Unlock()
// get host
host, ok := cluster.Nodes[hostName]
if !ok {
return nil, fmt.Errorf("%s not in cluster", hostName)
} else {
// aquire host lock to wait in case of a concurrent write
host.lock.Lock()
defer host.lock.Unlock()
go func() {
// aquire cluster lock
cluster.lock.Lock()
defer cluster.lock.Unlock()
// get host
host, ok := cluster.Nodes[hostName]
if !ok {
host_ch <- nil
err_ch <- fmt.Errorf("%s not in cluster", hostName)
} else {
// aquire host lock to wait in case of a concurrent write
host.lock.Lock()
defer host.lock.Unlock()
host_ch <- host
err_ch <- nil
}
}()
return <-host_ch, <-err_ch
return host, nil
}
}
// hard sync node
// returns error if the node could not be reached
func (cluster *Cluster) RebuildNode(hostName string) error {
err_ch := make(chan error)
host, err := cluster.pve.Node(hostName)
if err != nil && cluster.Nodes[hostName] == nil { // host is unreachable and did not exist previously
// return an error because we requested to sync a node that was not already in the cluster
return fmt.Errorf("error retrieving %s: %s", hostName, err.Error())
}
go func() {
host, err := cluster.pve.Node(hostName)
if err != nil && cluster.Nodes[hostName] == nil { // host is unreachable and did not exist previously
// return an error because we requested to sync a node that was not already in the cluster
err_ch <- fmt.Errorf("error retrieving %s: %s", hostName, err.Error())
}
// aquire lock on host, release on return
host.lock.Lock()
defer host.lock.Unlock()
// aquire lock on host, release on return
host.lock.Lock()
defer host.lock.Unlock()
wg, _ := errgroup.WithContext(context.Background())
if err != nil && cluster.Nodes[hostName] != nil { // host is unreachable and did exist previously
// assume the node is down or gone and delete from cluster
delete(cluster.Nodes, hostName)
err_ch <- nil
}
if err != nil && cluster.Nodes[hostName] != nil { // host is unreachable and did exist previously
// assume the node is down or gone and delete from cluster
delete(cluster.Nodes, hostName)
return nil
}
cluster.Nodes[hostName] = host
cluster.Nodes[hostName] = host
// get node's VMs
vms, err := host.VirtualMachines()
if err != nil {
err_ch <- err
// get node's VMs
vms, err := host.VirtualMachines()
if err != nil {
return err
}
for _, vmid := range vms {
}
for _, vmid := range vms {
wg.Go(func() error {
start := time.Now()
err := host.RebuildInstance(VM, vmid)
if err != nil { // if an error was encountered, continue and log the error
log.Printf("[ERR ] %s", err)
log.Printf("[ERR ] error encountered while syncing vm %s.%d: %s", hostName, vmid, err)
} else {
log.Printf("[INFO] successfully synced vm %s.%d", hostName, vmid)
log.Printf("[INFO] synced vm %s.%d in %d ms", hostName, vmid, time.Since(start).Milliseconds())
}
}
return err
})
}
// get node's CTs
cts, err := host.Containers()
if err != nil {
err_ch <- err
}
for _, vmid := range cts {
// get node's CTs
cts, err := host.Containers()
if err != nil {
return err
}
for _, vmid := range cts {
wg.Go(func() error {
start := time.Now()
err := host.RebuildInstance(CT, vmid)
if err != nil { // if an error was encountered, continue and log the error
log.Printf("[ERR ] %s", err)
log.Printf("[ERR ] error encountered while syncing ct %s.%d: %s", hostName, vmid, err)
} else {
log.Printf("[INFO] successfully synced ct %s.%d", hostName, vmid)
}
}
log.Printf("[INFO] synced ct %s.%d in %d ms", hostName, vmid, time.Since(start).Milliseconds())
// check node device reserved by iterating over each function, we will assume that a single reserved function means the device is also reserved
for _, device := range host.Devices {
reserved := false
for _, function := range device.Functions {
reserved = reserved || function.Reserved
}
device.Reserved = reserved
}
return err
})
}
err_ch <- nil
}()
return <-err_ch
err = wg.Wait()
if err != nil {
return err
}
// check node device reserved by iterating over each function, we will assume that a single reserved function means the device is also reserved
for _, device := range host.Devices {
reserved := false
for _, function := range device.Functions {
reserved = reserved || function.Reserved
}
device.Reserved = reserved
}
return nil
}
func (host *Node) GetInstance(vmid uint) (*Instance, error) {
instance_ch := make(chan *Instance)
err_ch := make(chan error)
go func() {
// aquire host lock
host.lock.Lock()
defer host.lock.Unlock()
// get instance
instance, ok := host.Instances[InstanceID(vmid)]
if !ok {
instance_ch <- nil
err_ch <- fmt.Errorf("vmid %d not in host %s", vmid, host.Name)
} else {
// aquire instance lock to wait in case of a concurrent write
instance.lock.Lock()
defer instance.lock.Unlock()
// aquire host lock
host.lock.Lock()
defer host.lock.Unlock()
// get instance
instance, ok := host.Instances[InstanceID(vmid)]
if !ok {
return nil, fmt.Errorf("vmid %d not in host %s", vmid, host.Name)
} else {
// aquire instance lock to wait in case of a concurrent write
instance.lock.Lock()
defer instance.lock.Unlock()
instance_ch <- instance
err_ch <- nil
}
}()
return <-instance_ch, <-err_ch
return instance, nil
}
}
// hard sync instance
// returns error if the instance could not be reached
func (host *Node) RebuildInstance(instancetype InstanceType, vmid uint) error {
err_ch := make(chan error)
instanceID := InstanceID(vmid)
var instance *Instance
var err error
switch instancetype {
case VM:
instance, err = host.VirtualMachine(vmid)
case CT:
instance, err = host.Container(vmid)
go func() {
instanceID := InstanceID(vmid)
var instance *Instance
var err error
switch instancetype {
case VM:
instance, err = host.VirtualMachine(vmid)
case CT:
instance, err = host.Container(vmid)
}
}
if err != nil && host.Instances[instanceID] == nil { // instance is unreachable and did not exist previously
// return an error because we requested to sync an instance that was not already in the cluster
return fmt.Errorf("error retrieving %s.%d: %s", host.Name, instanceID, err.Error())
}
if err != nil && host.Instances[instanceID] == nil { // instance is unreachable and did not exist previously
// return an error because we requested to sync an instance that was not already in the cluster
err_ch <- fmt.Errorf("error retrieving %s.%d: %s", host.Name, instanceID, err.Error())
}
// aquire lock on instance, release on return
instance.lock.Lock()
defer instance.lock.Unlock()
// aquire lock on instance, release on return
instance.lock.Lock()
defer instance.lock.Unlock()
wg, _ := errgroup.WithContext(context.Background())
if err != nil && host.Instances[instanceID] != nil { // host is unreachable and did exist previously
// assume the instance is gone and delete from cluster
delete(host.Instances, instanceID)
err_ch <- nil
}
if err != nil && host.Instances[instanceID] != nil { // host is unreachable and did exist previously
// assume the instance is gone and delete from cluster
delete(host.Instances, instanceID)
return nil
}
host.Instances[instanceID] = instance
host.Instances[instanceID] = instance
for volid := range instance.configDisks {
for volid := range instance.configDisks {
wg.Go(func() error {
err = instance.RebuildVolume(host, volid)
if err != nil {
err_ch <- err
log.Printf("[ERR ] error rebuilding volume %s: %s", volid, err)
}
}
return err
})
}
for netid := range instance.configNets {
for netid := range instance.configNets {
wg.Go(func() error {
err = instance.RebuildNet(host, netid)
if err != nil {
err_ch <- err
log.Printf("[ERR ] error rebuilding net %s: %s", netid, err)
return err
}
}
return err
})
}
for deviceid := range instance.configHostPCIs {
for deviceid := range instance.configHostPCIs {
wg.Go(func() error {
err = instance.RebuildDevice(host, deviceid)
if err != nil {
err_ch <- err
log.Printf("[ERR ] error rebuilding pci %s: %s", deviceid, err)
}
return err
})
}
err = wg.Wait()
if err != nil {
return err
}
if instance.Type == VM {
err = instance.RebuildBoot(host)
if err != nil {
log.Printf("[ERR ] error rebuilding boot: %s", err)
}
if instance.Type == VM {
err = instance.RebuildBoot(host)
if err != nil {
err_ch <- err
}
}
err_ch <- nil
}()
return <-err_ch
return err
} else {
return nil
}
}
func (instance *Instance) RebuildVolume(host *Node, volid string) error {