- Submit each task immediately for execution, instead of waiting for a scheduling thread to send all gathered tasks (up to that point) to the ExecutorService (and block until they are finished, before it can start again).
- Hold the Future of each submitted task to a synchronized-list to check the result of each task at a scheduled time.
- Reduce the cpu-time to assure the Service can shut down, by checking if there are "actively" and "about-to-be-executed" tasks, at the same time. Instead of having to rely on the additional checking of the "shutdown"-status of each worker to verify that no active task exist.
- Improve the threads' shutdown procedure.
- Refactor the "bulkImportReportID".
- Add the "bulk:" prefix in the provenance value, in the DB.
- Fix not using correctly the "Lists.partition()" method.
- Make sure the "bulkImportDir" is removed from the "bulkImportDirsUnderProcessing" Set, in case of an early-error.
- Fix the "numFailedSegments"-calculation.
- Improve some messages.
- Code polishing.
- Increase the "numOfThreadsPerBulkImportProcedure" to 6.
- Fix Bulk import not working from a second-level subdirectory; the report-subDirectory was not created.
- Fix not returning the bulk-import-report as "application/json".
- Add useful messages for missing parameters.
- Change the HTTP-method for the "bulkImportFullTexts" endpoint to "POST".
- Show a structured json-response for the "bulkImportFullTexts" endpoint.
- Fix uncommon date-format.
- Remove single quotes from json-report, since they are returned as bytes, not characters.
- Optimize the generation of the json-bulkImport-report.
After that enchantment, each worker could request multiple assignment-batches, before its previous batches were processed by the Controller. This means that for each batch that was processed, the Controller was deleting from the "assignment" table, all the assignments (-batches) delivered to the Worker that brought that batch, even though the "attempt" and "payload" records for the rest of the batches were not inserted in the DB yet. So in a new assignments-batch request, the same publications that were already under processing, were delivered to the same or other Workers.
Now, for each finished batch, only the assignments of that batch are deleted from the "assignment" table.
- Update the "StatsServiceImpl.getNumberOfPayloadsAggregatedByService()" to use the new table "payload_aggregated", instead of casting and checking the date of the records.
- Code polishing.
- Optimize performance of the code running right before the "postShutdownOrCancelRequestToWorker".
- Show which Workers have already shutdown and as a result a "postShutdownOrCancelRequestToWorker" will not be performed on them.
- Treat the "contains no visible files" situation as an error. In which case the assignments-data is presumed to not have been inserted to the database tables.
- Code polishing/cleanup.
- Improve startup speed, by using a faster remote server to get the host's machine public IP. This also reduces the risk of not being able to get the public IP at all.
- Fix the detection of a different IP for a known worker.
- Improve documentation.
- Store each worker's info in a hash-table, in order to efficiently know if we need to create new hdfs subdirectories. Also, this will help to issue "shutdown" requests to the workers in the future, as well as to know which worker has shutdown.
Lampros Smyrnaios