Skip to content

Beta 0.2.0

Latest
Compare
Choose a tag to compare
@hartfordfive hartfordfive released this 07 Jan 15:49
  • Added S3 state file storage option (Issue #2)
  • Fixed bug where nanosecond timestamp fields were not always received in scientific notiation, thus causing an error with the interface type conversion (Iusse #4)
  • Logs are now downloaded and stored in a gzip file, and then read sequentially in order to reduce memory requirement, which was previously higher due to all in-memory download and processing. (Issue #9)
  • Added new configuration options state_file_path, state_file_name, delete_logfile_after_processing, processed_events_buffer_size
  • Updated logic so that logs are downloaded immediately, without delay in the case where the process has stopped and the time elapsed is greater than the configured period.
  • Fixed Elasticsearch 5.x index template.
  • Explicitly closed log files handles upon completion as the too many files open error was begining to occur after the process was running for over a few days (Issue #11)
  • Added BuildMapStr function which builds the final event to be sent. This ensures that fields with types such as ip will simply be ommitted if they are an empty string, otherwise this used to cause a mapping exception.
  • Included the zone_tag in the state file name so that each Cloudflare zone will have its own state file.