Documentation
¶
Index ¶
Constants ¶
const ( // ContainerPackagesDir represents the default location of installed BOSH packages ContainerPackagesDir = "/var/vcap/packages" // ContainerSourceDir is the directory in which the source code will reside when we // compile them. We will add a volume mount there in the container to work around // issues with AUFS not emulating a normal filesystem correctly. ContainerSourceDir = "/var/vcap/source" )
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Compilator ¶
type Compilator struct {
// contains filtered or unexported fields
}
Compilator represents the BOSH compiler
func NewDockerCompilator ¶
func NewDockerCompilator( dockerManager *docker.ImageManager, hostWorkDir string, metricsPath string, stemcellImageName string, baseType string, fissileVersion string, dockerNetworkMode string, keepContainer bool, ui *termui.UI, grapher util.ModelGrapher, packageStorage *PackageStorage, streamPackages bool, ) (*Compilator, error)
NewDockerCompilator will create an instance of the Compilator using docker
func NewMountNSCompilator ¶
func NewMountNSCompilator( hostWorkDir string, metricsPath string, stemcellImageName string, baseType string, fissileVersion string, ui *termui.UI, grapher util.ModelGrapher, packageStorage *PackageStorage, ) (*Compilator, error)
NewMountNSCompilator will create an instance of the Compilator using a mount namespace (Linux only)
func (*Compilator) Compile ¶
func (c *Compilator) Compile(workerCount int, releases []*model.Release, instanceGroups model.InstanceGroups, verbose bool) error
Compile concurrency works like this: 1 routine producing (todoCh<-) <=> Compile() itself n workers consuming (<-todoCh) <=> compileJob.Run()'s 1 synchronizer consuming EXACTLY 1 <-doneCh for every <-todoCh <=> Compile() again.
Dependencies:
- Packages with the least dependencies are queued first.
- Workers wait for their dependencies by waiting on a map of broadcasting channels that are closed by the synchronizer when something is done compiling successfully ==> c.signalDependencies [<fingerprint>]
In the event of an error:
workers will try to bail out of waiting on <-todo or <-c.signalDependencies[<fingerprint>] early if it finds the killCh has been activated. There is a "race" here to see if the synchronizer will drain <-todoCh or if they will select on <-killCh before <-todoCh. In the worst case, extra packages will be compiled by each active worker. See (**), (xx)
Note that jobs without dependencies ignore the kill signal. See (xx).
- synchronizer will greedily drain the <-todoCh to starve the workers out and won't wait for the <-doneCh for the N packages it drained.
type DownloadProgressEvent ¶
type DownloadProgressEvent = func(progressPercentage float64)
DownloadProgressEvent represents a delegate for updating progress when downloading
type PackageStorage ¶
type PackageStorage struct {
Kind string
Config stow.Config
CompilationWorkDir string
ImageName string
ReadOnly bool
// contains filtered or unexported fields
}
PackageStorage represents a compiled BOSH package location
func NewPackageStorage ¶
func NewPackageStorage(kind string, readOnlyMode bool, config stow.Config, compilationWorkDir string, containerPath string, stemcellImageName string) (p *PackageStorage, err error)
NewPackageStorage creates a new PackageStorage instance
func NewPackageStorageFromConfig ¶
func NewPackageStorageFromConfig(configFilePath, compilationWorkDir, stemcellImageName string) (*PackageStorage, error)
NewPackageStorageFromConfig creates a new PackageStorage based on a configuration file
func (*PackageStorage) Download ¶
func (p *PackageStorage) Download(pack *model.Package, progressEvent DownloadProgressEvent) error
Download downloads a package from the configured cache
Source Files