// checkExistingFile checks if a harvester has to be started for a already known file
// For existing files the following options exist:
// * Last reading position is 0, no harvester has to be started as old harvester probably still busy
// * The old known modification time is older then the current one. Start at last known position
// * The new file is not the same as the old file, means file was renamed
// ** New file is actually really a new file, start a new harvester
// ** Renamed file has a state, continue there
func (p *Prospector) checkExistingFile(newinfo *harvester.FileStat, newFile *input.File, oldFile *input.File, file string, output chan *input.FileEvent) {
logp.Debug("prospector", "Update existing file for harvesting: %s", file)
h, err := harvester.NewHarvester(
p.ProspectorConfig, &p.ProspectorConfig.Harvester,
file, newinfo, output)
if err != nil {
logp.Err("Error initializing harvester: %v", err)
return
}
if !oldFile.IsSameFile(newFile) {
if previousFile, err := p.getPreviousFile(file, newinfo.Fileinfo); err == nil {
// This file was renamed from another file we know - link the same harvester channel as the old file
logp.Debug("prospector", "File rename was detected: %s -> %s", previousFile, file)
logp.Debug("prospector", "Launching harvester on renamed file: %s", file)
lastinfo := p.prospectorList[previousFile]
newinfo.Continue(&lastinfo)
} else {
// File is not the same file we saw previously, it must have rotated and is a new file
logp.Debug("prospector", "Launching harvester on rotated file: %s", file)
// Forget about the previous harvester and let it continue on the old file - so start a new channel to use with the new harvester
newinfo.Ignore()
// Start a new harvester on the path
h.Start()
}
// Keep the old file in missingFiles so we don't rescan it if it was renamed and we've not yet reached the new filename
// We only need to keep it for the remainder of this iteration then we can assume it was deleted and forget about it
p.missingFiles[file] = oldFile.FileInfo
} else if newinfo.Finished() && oldFile.FileInfo.ModTime() != newinfo.Fileinfo.ModTime() {
// Resume harvesting of an old file we've stopped harvesting from
logp.Debug("prospector", "Resuming harvester on an old file that was just modified: %s", file)
// Start a harvester on the path; an old file was just modified and it doesn't have a harvester
// The offset to continue from will be stored in the harvester channel - so take that to use and also clear the channel
h.Offset = <-newinfo.Return
h.Start()
} else {
logp.Debug("prospector", "Not harvesting, file didn't change: %s", file)
}
}
// checkExistingFile checks if a harvester has to be started for a already known file
// For existing files the following options exist:
// * Last reading position is 0, no harvester has to be started as old harvester probably still busy
// * The old known modification time is older then the current one. Start at last known position
// * The new file is not the same as the old file, means file was renamed
// ** New file is actually really a new file, start a new harvester
// ** Renamed file has a state, continue there
func (p *Prospector) checkExistingFile(newinfo *ProspectorFileStat, newFile *input.File, oldFile *input.File, file string, output chan *input.FileEvent) {
logp.Debug("prospector", "Update existing file for harvesting: %s", file)
h := &harvester.Harvester{
Path: file,
FileConfig: p.FileConfig,
FinishChan: newinfo.Harvester,
SpoolerChan: output,
}
if !oldFile.IsSameFile(newFile) {
if previousFile := p.isFileRenamed(file, newinfo.Fileinfo); previousFile != "" {
// This file was renamed from another file we know - link the same harvester channel as the old file
logp.Debug("prospector", "File rename was detected: %s -> %s", previousFile, file)
logp.Debug("prospector", "Launching harvester on renamed file: %s", file)
newinfo.Harvester = p.prospectorList[previousFile].Harvester
} else {
// File is not the same file we saw previously, it must have rotated and is a new file
logp.Debug("prospector", "Launching harvester on rotated file: %s", file)
// Forget about the previous harvester and let it continue on the old file - so start a new channel to use with the new harvester
newinfo.Harvester = make(chan int64, 1)
// Start a new harvester on the path
h.Start()
}
// Keep the old file in missingFiles so we don't rescan it if it was renamed and we've not yet reached the new filename
// We only need to keep it for the remainder of this iteration then we can assume it was deleted and forget about it
p.missingFiles[file] = oldFile.FileInfo
} else if len(newinfo.Harvester) != 0 && oldFile.FileInfo.ModTime() != newinfo.Fileinfo.ModTime() {
// Resume harvesting of an old file we've stopped harvesting from
logp.Debug("prospector", "Resuming harvester on an old file that was just modified: %s", file)
// Start a harvester on the path; an old file was just modified and it doesn't have a harvester
// The offset to continue from will be stored in the harvester channel - so take that to use and also clear the channel
h.Offset = <-newinfo.Harvester
h.Start()
} else {
logp.Debug("prospector", "Not harvesting, file didn't change: %s", file)
}
}
