diff --git a/beat/web/backend/models.py b/beat/web/backend/models.py index 16f712127493866472b442426003603d62619655..609abed7941371095b29b93f27a4210d98450268 100644 --- a/beat/web/backend/models.py +++ b/beat/web/backend/models.py @@ -440,9 +440,9 @@ class Worker(models.Model): else: # process went away without any apparent reason with transaction.atomic(): message = "Split %d/%d running at worker `%s' for " \ - "block `%s' of experiment `%s' finished without any " \ - "apparent reason. Checking-out job split at " \ - "database by force..." % (j.split_index+1, + "block `%s' of experiment `%s' finished before " \ + "even starting. Force-cancelling job split at " \ + "database..." % (j.split_index+1, j.job.block.required_slots, self, j.job.block.name, @@ -1485,6 +1485,9 @@ class JobSplit(models.Model): self, os.getpid(), traceback.format_exc()) self.signal_io_error() if self.cache_errors > settings.MAXIMUM_IO_ERRORS: + logger.info("Split `%s' reached the maximum number of IO " \ + "errors (%d > %d). Force failing this split." % \ + (self, self.cache_errors, settings.MAXIMUM_IO_ERRORS)) self.try_end(Result(status=1, usrerr=settings.DEFAULT_USER_ERROR, syserr=traceback.format_exc(),))