import (
"fmt"
+ "io/ioutil"
"os"
"strings"
type deprCluster struct {
RequestLimits deprRequestLimits
- NodeProfiles map[string]arvados.NodeProfile
+ NodeProfiles map[string]nodeProfile
}
type deprecatedConfig struct {
Clusters map[string]deprCluster
}
-func applyDeprecatedConfig(cfg *arvados.Config, configdata []byte, log logger) error {
+type nodeProfile struct {
+ Controller systemServiceInstance `json:"arvados-controller"`
+ Health systemServiceInstance `json:"arvados-health"`
+ Keepbalance systemServiceInstance `json:"keep-balance"`
+ Keepproxy systemServiceInstance `json:"keepproxy"`
+ Keepstore systemServiceInstance `json:"keepstore"`
+ Keepweb systemServiceInstance `json:"keep-web"`
+ Nodemanager systemServiceInstance `json:"arvados-node-manager"`
+ DispatchCloud systemServiceInstance `json:"arvados-dispatch-cloud"`
+ RailsAPI systemServiceInstance `json:"arvados-api-server"`
+ Websocket systemServiceInstance `json:"arvados-ws"`
+ Workbench1 systemServiceInstance `json:"arvados-workbench"`
+}
+
+type systemServiceInstance struct {
+ Listen string
+ TLS bool
+ Insecure bool
+}
+
+func (ldr *Loader) applyDeprecatedConfig(cfg *arvados.Config) error {
var dc deprecatedConfig
- err := yaml.Unmarshal(configdata, &dc)
+ err := yaml.Unmarshal(ldr.configdata, &dc)
if err != nil {
return err
}
for name, np := range dcluster.NodeProfiles {
if name == "*" || name == os.Getenv("ARVADOS_NODE_PROFILE") || name == hostname {
name = "localhost"
- } else if log != nil {
- log.Warnf("overriding Clusters.%s.Services using Clusters.%s.NodeProfiles.%s (guessing %q is a hostname)", id, id, name, name)
+ } else if ldr.Logger != nil {
+ ldr.Logger.Warnf("overriding Clusters.%s.Services using Clusters.%s.NodeProfiles.%s (guessing %q is a hostname)", id, id, name, name)
}
applyDeprecatedNodeProfile(name, np.RailsAPI, &cluster.Services.RailsAPI)
applyDeprecatedNodeProfile(name, np.Controller, &cluster.Services.Controller)
return nil
}
-func applyDeprecatedNodeProfile(hostname string, ssi arvados.SystemServiceInstance, svc *arvados.Service) {
+func applyDeprecatedNodeProfile(hostname string, ssi systemServiceInstance, svc *arvados.Service) {
scheme := "https"
if !ssi.TLS {
scheme = "http"
}
svc.InternalURLs[arvados.URL{Scheme: scheme, Host: host}] = arvados.ServiceInstance{}
}
+
+const defaultKeepstoreConfigPath = "/etc/arvados/keepstore/keepstore.yml"
+
+type oldKeepstoreConfig struct {
+ Debug *bool
+}
+
+func (ldr *Loader) loadOldConfigHelper(component, path string, target interface{}) error {
+ if path == "" {
+ return nil
+ }
+ buf, err := ioutil.ReadFile(path)
+ if err != nil {
+ return err
+ }
+
+ ldr.Logger.Warnf("you should remove the legacy %v config file (%s) after migrating all config keys to the cluster configuration file (%s)", component, path, ldr.Path)
+
+ err = yaml.Unmarshal(buf, target)
+ if err != nil {
+ return fmt.Errorf("%s: %s", path, err)
+ }
+ return nil
+}
+
+// update config using values from an old-style keepstore config file.
+func (ldr *Loader) loadOldKeepstoreConfig(cfg *arvados.Config) error {
+ if ldr.KeepstorePath == "" {
+ return nil
+ }
+ var oc oldKeepstoreConfig
+ err := ldr.loadOldConfigHelper("keepstore", ldr.KeepstorePath, &oc)
+ if os.IsNotExist(err) && (ldr.KeepstorePath == defaultKeepstoreConfigPath) {
+ return nil
+ } else if err != nil {
+ return err
+ }
+
+ cluster, err := cfg.GetCluster("")
+ if err != nil {
+ return err
+ }
+
+ if v := oc.Debug; v == nil {
+ } else if *v && cluster.SystemLogs.LogLevel != "debug" {
+ cluster.SystemLogs.LogLevel = "debug"
+ } else if !*v && cluster.SystemLogs.LogLevel != "info" {
+ cluster.SystemLogs.LogLevel = "info"
+ }
+
+ cfg.Clusters[cluster.ClusterID] = *cluster
+ return nil
+}
+
+type oldCrunchDispatchSlurmConfig struct {
+ Client *arvados.Client
+
+ SbatchArguments *[]string
+ PollPeriod *arvados.Duration
+ PrioritySpread *int64
+
+ // crunch-run command to invoke. The container UUID will be
+ // appended. If nil, []string{"crunch-run"} will be used.
+ //
+ // Example: []string{"crunch-run", "--cgroup-parent-subsystem=memory"}
+ CrunchRunCommand *[]string
+
+ // Extra RAM to reserve (in Bytes) for SLURM job, in addition
+ // to the amount specified in the container's RuntimeConstraints
+ ReserveExtraRAM *int64
+
+ // Minimum time between two attempts to run the same container
+ MinRetryPeriod *arvados.Duration
+
+ // Batch size for container queries
+ BatchSize *int64
+}
+
+const defaultCrunchDispatchSlurmConfigPath = "/etc/arvados/crunch-dispatch-slurm/crunch-dispatch-slurm.yml"
+
+func loadOldClientConfig(cluster *arvados.Cluster, client *arvados.Client) {
+ if client == nil {
+ return
+ }
+ if client.APIHost != "" {
+ cluster.Services.Controller.ExternalURL.Host = client.APIHost
+ }
+ if client.Scheme != "" {
+ cluster.Services.Controller.ExternalURL.Scheme = client.Scheme
+ } else {
+ cluster.Services.Controller.ExternalURL.Scheme = "https"
+ }
+ if client.AuthToken != "" {
+ cluster.SystemRootToken = client.AuthToken
+ }
+ cluster.TLS.Insecure = client.Insecure
+ ks := ""
+ for i, u := range client.KeepServiceURIs {
+ if i > 0 {
+ ks += " "
+ }
+ ks += u
+ }
+ cluster.Containers.SLURM.SbatchEnvironmentVariables = map[string]string{"ARVADOS_KEEP_SERVICES": ks}
+}
+
+// update config using values from an crunch-dispatch-slurm config file.
+func (ldr *Loader) loadOldCrunchDispatchSlurmConfig(cfg *arvados.Config) error {
+ if ldr.CrunchDispatchSlurmPath == "" {
+ return nil
+ }
+ var oc oldCrunchDispatchSlurmConfig
+ err := ldr.loadOldConfigHelper("crunch-dispatch-slurm", ldr.CrunchDispatchSlurmPath, &oc)
+ if os.IsNotExist(err) && (ldr.CrunchDispatchSlurmPath == defaultCrunchDispatchSlurmConfigPath) {
+ return nil
+ } else if err != nil {
+ return err
+ }
+
+ cluster, err := cfg.GetCluster("")
+ if err != nil {
+ return err
+ }
+
+ loadOldClientConfig(cluster, oc.Client)
+
+ if oc.SbatchArguments != nil {
+ cluster.Containers.SLURM.SbatchArgumentsList = *oc.SbatchArguments
+ }
+ if oc.PollPeriod != nil {
+ cluster.Containers.CloudVMs.PollInterval = *oc.PollPeriod
+ }
+ if oc.PrioritySpread != nil {
+ cluster.Containers.SLURM.PrioritySpread = *oc.PrioritySpread
+ }
+ if oc.CrunchRunCommand != nil {
+ if len(*oc.CrunchRunCommand) >= 1 {
+ cluster.Containers.CrunchRunCommand = (*oc.CrunchRunCommand)[0]
+ }
+ if len(*oc.CrunchRunCommand) >= 2 {
+ cluster.Containers.CrunchRunArgumentsList = (*oc.CrunchRunCommand)[1:]
+ }
+ }
+ if oc.ReserveExtraRAM != nil {
+ cluster.Containers.ReserveExtraRAM = arvados.ByteSize(*oc.ReserveExtraRAM)
+ }
+ if oc.MinRetryPeriod != nil {
+ cluster.Containers.MinRetryPeriod = *oc.MinRetryPeriod
+ }
+ if oc.BatchSize != nil {
+ cluster.API.MaxItemsPerResponse = int(*oc.BatchSize)
+ }
+
+ cfg.Clusters[cluster.ClusterID] = *cluster
+ return nil
+}
+
+type oldWsConfig struct {
+ Client *arvados.Client
+ Postgres *arvados.PostgreSQLConnection
+ PostgresPool *int
+ Listen *string
+ LogLevel *string
+ LogFormat *string
+
+ PingTimeout *arvados.Duration
+ ClientEventQueue *int
+ ServerEventQueue *int
+
+ ManagementToken *string
+}
+
+const defaultWebsocketConfigPath = "/etc/arvados/ws/ws.yml"
+
+// update config using values from an crunch-dispatch-slurm config file.
+func (ldr *Loader) loadOldWebsocketConfig(cfg *arvados.Config) error {
+ if ldr.WebsocketPath == "" {
+ return nil
+ }
+ var oc oldWsConfig
+ err := ldr.loadOldConfigHelper("arvados-ws", ldr.WebsocketPath, &oc)
+ if os.IsNotExist(err) && ldr.WebsocketPath == defaultWebsocketConfigPath {
+ return nil
+ } else if err != nil {
+ return err
+ }
+
+ cluster, err := cfg.GetCluster("")
+ if err != nil {
+ return err
+ }
+
+ loadOldClientConfig(cluster, oc.Client)
+
+ if oc.Postgres != nil {
+ cluster.PostgreSQL.Connection = *oc.Postgres
+ }
+ if oc.PostgresPool != nil {
+ cluster.PostgreSQL.ConnectionPool = *oc.PostgresPool
+ }
+ if oc.Listen != nil {
+ cluster.Services.Websocket.InternalURLs[arvados.URL{Host: *oc.Listen}] = arvados.ServiceInstance{}
+ }
+ if oc.LogLevel != nil {
+ cluster.SystemLogs.LogLevel = *oc.LogLevel
+ }
+ if oc.LogFormat != nil {
+ cluster.SystemLogs.Format = *oc.LogFormat
+ }
+ if oc.PingTimeout != nil {
+ cluster.API.SendTimeout = *oc.PingTimeout
+ }
+ if oc.ClientEventQueue != nil {
+ cluster.API.WebsocketClientEventQueue = *oc.ClientEventQueue
+ }
+ if oc.ServerEventQueue != nil {
+ cluster.API.WebsocketServerEventQueue = *oc.ServerEventQueue
+ }
+ if oc.ManagementToken != nil {
+ cluster.ManagementToken = *oc.ManagementToken
+ }
+
+ cfg.Clusters[cluster.ClusterID] = *cluster
+ return nil
+}
+
+type oldKeepProxyConfig struct {
+ Client *arvados.Client
+ Listen *string
+ DisableGet *bool
+ DisablePut *bool
+ DefaultReplicas *int
+ Timeout *arvados.Duration
+ PIDFile *string
+ Debug *bool
+ ManagementToken *string
+}
+
+const defaultKeepproxyConfigPath = "/etc/arvados/keepproxy/keepproxy.yml"
+
+func (ldr *Loader) loadOldKeepproxyConfig(cfg *arvados.Config) error {
+ if ldr.KeepproxyPath == "" {
+ return nil
+ }
+ var oc oldKeepProxyConfig
+ err := ldr.loadOldConfigHelper("keepproxy", ldr.KeepproxyPath, &oc)
+ if os.IsNotExist(err) && ldr.KeepproxyPath == defaultKeepproxyConfigPath {
+ return nil
+ } else if err != nil {
+ return err
+ }
+
+ cluster, err := cfg.GetCluster("")
+ if err != nil {
+ return err
+ }
+
+ loadOldClientConfig(cluster, oc.Client)
+
+ if oc.Listen != nil {
+ cluster.Services.Keepproxy.InternalURLs[arvados.URL{Host: *oc.Listen}] = arvados.ServiceInstance{}
+ }
+ if oc.DefaultReplicas != nil {
+ cluster.Collections.DefaultReplication = *oc.DefaultReplicas
+ }
+ if oc.Timeout != nil {
+ cluster.API.KeepServiceRequestTimeout = *oc.Timeout
+ }
+ if oc.Debug != nil {
+ if *oc.Debug && cluster.SystemLogs.LogLevel != "debug" {
+ cluster.SystemLogs.LogLevel = "debug"
+ } else if !*oc.Debug && cluster.SystemLogs.LogLevel != "info" {
+ cluster.SystemLogs.LogLevel = "info"
+ }
+ }
+ if oc.ManagementToken != nil {
+ cluster.ManagementToken = *oc.ManagementToken
+ }
+
+ // The following legacy options are no longer supported. If they are set to
+ // true or PIDFile has a value, error out and notify the user
+ unsupportedEntry := func(cfgEntry string) error {
+ return fmt.Errorf("the keepproxy %s configuration option is no longer supported, please remove it from your configuration file", cfgEntry)
+ }
+ if oc.DisableGet != nil && *oc.DisableGet {
+ return unsupportedEntry("DisableGet")
+ }
+ if oc.DisablePut != nil && *oc.DisablePut {
+ return unsupportedEntry("DisablePut")
+ }
+ if oc.PIDFile != nil && *oc.PIDFile != "" {
+ return unsupportedEntry("PIDFile")
+ }
+
+ cfg.Clusters[cluster.ClusterID] = *cluster
+ return nil
+}