// 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 ProspectorLog) checkExistingFile(h *harvester.Harvester, newFile *input.File, oldFile *input.File) { logp.Debug("prospector", "Update existing file for harvesting: %s", h.Path) // We assume it is the same file, but it wasn't if !oldFile.IsSameFile(newFile) { logp.Debug("prospector", "File previously found: %s", h.Path) if previousFile, err := p.getPreviousFile(h.Path, h.Stat.Fileinfo); err == nil { p.continueExistingFile(h, previousFile) } 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", h.Path) // Forget about the previous harvester and let it continue on the old file - so start a new channel to use with the new harvester h.Stat.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[h.Path] = oldFile.FileInfo } else if h.Stat.Finished() && oldFile.FileInfo.ModTime() != h.Stat.Fileinfo.ModTime() { // Resume harvesting of an old file we've stopped harvesting from // Start a harvester on the path; a 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 p.resumeHarvesting(h, <-h.Stat.Return) } else { logp.Debug("prospector", "Not harvesting, file didn't change: %s", h.Path) } }
// Continue reading on an existing file. // The given file was renamed from another file we know -> The same harvester channel is linked as the old file // The file param is only used for logging func (p *ProspectorLog) continueExistingFile(h *harvester.Harvester, previousFile string) { logp.Debug("prospector", "Launching harvester on renamed file. File rename was detected: %s -> %s", previousFile, h.Path) lastinfo := p.harvesterStats[previousFile] h.Stat.Continue(&lastinfo) // Update state because of file rotation h.UpdateState() }
// Start / resume harvester with a predefined offset func (p *ProspectorLog) resumeHarvesting(h *harvester.Harvester, offset int64) { logp.Debug("prospector", "Start / resuming harvester of file: %s", h.Path) h.SetOffset(offset) h.Start() }