retry-log

retry-log

Process events in a log. Retry them a number of times. Then file them into failed.log or fixed.log

var processLog = require('retry-log')
var retryJob = require('your magic function that performs work')

var log = "./problems.log"

retry(log,function(data,cb){
  retryJob(data,cb)
},function(){
  console.log('all done processing')
})

API

• retry(options,handler,done)
  • returns undefined
  • options
    • string, log file name
    • object
      • tries: 3, the number of times each job should be tried.
      • log: required. the log file to process
      • failedLog: the log file to write failed tasks
      • fixedLog: the log file to write fixed tasks
      • pid: optional, the pid of the process that writes the source log file.
      • rotateSignal: SIGHUP. if pid is provided this signal is delivered to the process. It is expected that it get a new handle to the log file because it will be moved and unlinked as part of managing the retries.

Files

this manages marshalling events in a log file to either a failed log or a fixed log

the log file.

• the log file must contain newline delimited json
• this module reserves 2 keys on this object. "tries" and "attempts"
  • "tries" is a number that incrments with each attempt
  • "attempts" is an array of {error:,data:} objects with the results of each attempt.

log life cycle

• create job id (monotonic-timestamp)
• rename log file to logname+".data_"+jobId
• if pid is provided, signal pid and blindly trust that it releases the file descriptor
• create a read stream with fs-readstream-many which is exactly the same ascat $logname.data_*
• as each line is processed based on the result write to one of 3 logs.
  • if the job passes the line is appended to fixed.log
  • if the job fails and there are no more chances the line is appended to failed.log
  • if the job failed but still has more chances. the line gets added to logname+".tmp_"+jobId
• after all jobs are complete if any lines have been written to the new retry log move logname+".tmp_"+jobId to logname+".data_"+jobId+"_result"
• unlink all other logname+".data_* logs that were processed