b94c35c66e- Fix double active "@Scheduled" annotation for the "ScheduledTasks.updatePrometheusMetrics()" method. - Code polishing.Lampros Smyrnaios2023-07-13 18:32:45 +0300
8dfb58ee63Avoid assigning the same publications multiple times to the Workers, after the recent "parallelization enchantment". 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.Lampros Smyrnaios2023-07-11 17:27:23 +0300
d5c139c410Handle the case where the "stats"-queries are executed while some table of the DB are in a "merge" state. In this case, the queries fail and the Controller retries up to 10 times.Lampros Smyrnaios2023-07-06 18:29:13 +0300
e8644cb64f- Optimize the "insertAssignmentsQuery". - Add documentation about the Prometheus Metrics, in README. - Update Dependencies. - Code polishing.Lampros Smyrnaios2023-07-05 17:10:30 +0300
a89abe3f2fPrioritize the publications, which are specified inside the "publication_boost" table, according to their "boost-level".Lampros Smyrnaios2023-06-29 12:32:06 +0300
4c3e2e6b6e- Fix not using actual the "currentAssignmentsBatch" of the workerReport itself, when creating the parquetFileNames and when reporting to the user the initialization of the "addition of the workerReport". - Code polishing.Lampros Smyrnaios2023-06-27 16:08:01 +0300
0f4b63c4a9Expose the following statistics as prometheus-metrics and create/update a stats-endpoint for each one: - "numOfPayloadsAggregatedByServiceThroughCrawling" - "numOfPayloadsAggregatedByServiceThroughBulkImport" - "numOfPayloadsAggregatedByService" - "numOfLegacyPayloads" - "numOfRecordsInspectedByServiceThroughCrawling" (renamed from "numOfInspectedRecords")Lampros Smyrnaios2023-06-23 15:22:26 +0300
b6b1cb08b9Add instructions on how to run the Prometheus and Grafana docker-containers alongside the UrlsController, by using the same script.Lampros Smyrnaios2023-06-23 14:52:07 +0300
d52601e819- Add handling for the "EmptyResultDataAccessException" in "UrlsServiceImpl.getAssignments()", which is thrown when no assignments are returned from the query. - Code polishing.Lampros Smyrnaios2023-06-22 12:39:11 +0300
b9712bed85- Expose the "numOfAllPayloads" and "numOfInspectedRecords" DB-stats to Prometheus, by using a scheduling task to request the numbers from the DB, every 6 hours. - Update the "StatsServiceImpl.getNumberOfPayloadsAggregatedByService()" to use the new table "payload_aggregated", instead of casting and checking the date of the records. - Code polishing.Lampros Smyrnaios2023-06-19 14:42:00 +0300
798fa09d68- Identify and handle a possible Worker-crash, in "UrlsServiceImpl.postReportResultToWorker()". - Add/Improve some log messages. - Update and cleanup dependencies. - Code polishing.Lampros Smyrnaios2023-06-15 23:19:36 +0300
88a74b2c41Add support for all private addresses, defined in "RFC 1918" standard. This fixes the issue of discarding some "shutdownService" requests due to coming from different local private addresses, when the Controller was run inside a docker container.Lampros Smyrnaios2023-06-15 13:26:27 +0300
c37f157f51Split the full-texts-batch's main handling-code to two separate methods, which can be used in parallel by two threads, in the future.Lampros Smyrnaios2023-06-14 17:16:38 +0300
e2776c50d0- Optimize the "WorkerReportResult" and the "ShutdownWorker" requests. - Improve documentation.Lampros Smyrnaios2023-06-10 02:31:57 +0300
0b1ab5b991Attempt to recover from serious failures, by individualizing the error-handling for each of the "table-merging" queries.Lampros Smyrnaios2023-06-10 02:28:02 +0300
6669dc61bf- Increase the initialDelay for the "checkIfServiceIsReadyForShutdown" scheduled-task, in production, to 10 minutes. - Code polishing.Lampros Smyrnaios2023-06-06 16:49:53 +0300
5d99a4be5d- Add the Shutdown Service API documentation. - Improve the BulkImport API documentation. - Fix markdown in README. - Update the app's version.Lampros Smyrnaios2023-06-06 16:18:38 +0300
54685bbe9a- Avoid sending "cancelShutdown" requests to already shutDown Workers. - 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.Lampros Smyrnaios2023-05-29 13:41:37 +0300
03bf4294b8- Add documentation about the "BulkImport API" in the README. - Fix a link in README. - Update dependencies.Lampros Smyrnaios2023-05-29 12:13:39 +0300
74ff31fc64- Show the workerIPs in the logs. - Rename the "FullTexts"-files to "BulkImport".Lampros Smyrnaios2023-05-29 12:12:08 +0300
3988eb3a48- Use a separate HDFS sub-dir for every assignments-batch, in order to avoid any disrruptancies from multiple threads moving parquet-files from the same sub-dir. Multiple batches from the same worker may be processed at the same time. These sub-dirs are deleted afterwards. - 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.Lampros Smyrnaios2023-05-27 02:36:05 +0300
02cee097d4Fix an issue, which could cause some background jobs to be executed more than 1 times. The previously executed jobs were not deleted from the global list fast enough, and they would be selected again, in case they were not finished before the scheduler started again.Lampros Smyrnaios2023-05-26 13:08:00 +0300
2b50e08bf6- Handle the case, were multiple threads may load the same HDFS directory to a database table, thus causing the "directory contains no visible files"-SQLException. - Improve the values of the delays for some scheduledTasks. - Improve elapsed time precision for the "lastAccessedOn" metadata of the workerReports. - Code polishing.Lampros Smyrnaios2023-05-25 00:34:36 +0300
164245cb53- Automatically delete the unsuccessful WorkerReports, which are more than 7 days old. - Optimize the Service's startup speed, by setting "initialDelays" to the scheduled tasks. - Optimize documentation.Lampros Smyrnaios2023-05-24 16:59:42 +0300
8b5f143b0aPlace the "workerReports" and the "bulkImportReports" dirs inside the "reports" parent-directory.Lampros Smyrnaios2023-05-24 14:10:57 +0300
cd1fb0af88- Process the WorkerReports in background Jobs and post the reportResults to the Workers. - Save the workerReports to json files, until they are processed successfully. - Show some custom metrics in prometheus.Lampros Smyrnaios2023-05-24 13:52:28 +0300
0ea3e2de24Add the "shutdownService" and "cancelShutdownService" endpoints. The Controller sends the related requests to the Workers and shutdowns gracefully, after all workers have shutdown.Lampros Smyrnaios2023-05-24 13:42:29 +0300
c2a1b96069- Rename the mounted "mnt/bulkImport/" directory to "/mnt/bulk_import/". - Increase the "awaitTermination" timeout for the ExecutorService to 2 minutes.Lampros Smyrnaios2023-05-23 21:09:34 +0300
c7bfd75973- Add the "getWorkersInfo" endpoint. - 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.Lampros Smyrnaios2023-05-23 14:57:15 +0300
5f75b48e95- Increase the "read-timeout" when searching for the host's machine public-IP. - Update dependencies. - Code polishing.Lampros Smyrnaios2023-05-22 21:33:02 +0300
0ab6bae93a- Optimize the json-conversion of the "BulkImportReport". - Code polishing.Lampros Smyrnaios2023-05-18 17:30:40 +0300
f7f919cee1- Make sure we set the "hasShutdown" to "false", for each known worker which was restarted. - Fix markdown of urls in prometheus' readme.Lampros Smyrnaios2023-05-16 12:24:14 +0300
b499209ce3- Move the Prometheus and grafana configuration in a dedicated directory and docker-compose file. - Add documentation about setting-up prometheus and grafana.Lampros Smyrnaios2023-05-15 18:52:31 +0300
f51a34138f- Use separate HDFS subdirectories for each worker in order to avoid seeing exceptions about "empty hdfs directory" when "loading" data to the database, because one worker has loaded data generated by multiple workers (since we use only 1 load operation for multiple parquet files). - 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 Smyrnaios2023-05-15 13:12:20 +0300
9412391903- In test-environment mode, check for already existing file-hashes only in the "payload_aggregated" table, instead of the whole "payload" view. This way the investigation for false-positive docUrls is easier, as we avoid checking against the millions of "legacy" payloads. - Improve performance in production, by not creating the string objects for "trace"-logs.Lampros Smyrnaios2023-05-15 12:44:16 +0300
8381df70c6- Improve performance of uploading parquet-files to HDFS. - Add some logs. - Code polishing.Lampros Smyrnaios2023-05-11 19:40:48 +0300
992d4ffd5e- Add the time-zone in the logs. - Change some log-levels to "trace", although most of them are still disabled.Lampros Smyrnaios2023-05-11 03:10:53 +0300
42b93e9429- Add the "getNumberOfAllDistinctFullTexts" stats-endpoint. - Add TODOs for more stats endpoints. - Code polishing.Lampros Smyrnaios2023-05-04 15:48:49 +0300
b3196376ebFix a bug, which caused the full-text files to never close.Lampros Smyrnaios2023-05-04 13:03:28 +0300
1b14a7e554- Add profiles to docker-services to selectively run the additional "Prometheus" and "Grafana" services or not. - Update Gradle.Lampros Smyrnaios2023-04-22 16:50:33 +0300
c2b17163cdAutomatically show the Controller's logs after the docker-container starts running and the status is shown.Lampros Smyrnaios2023-04-11 11:59:10 +0300
4dc34429f8- Increase the waiting-time before checking the docker containers' status, in order to catch configuration-crashes. - Code polishing.Lampros Smyrnaios2023-04-10 22:28:53 +0300
c39fef2654Upgrade payload-table to payload-view which consists of three separate payload tables: "payload_legacy", "payload_aggregated" and "payload_bulk_import".Lampros Smyrnaios2023-04-10 15:55:50 +0300
484cf5cefc- Avoid requesting the remaining full-text batches in case the Worker returns a 5XX error in one of the batches. - Add nullability-checks for "datasourceId" and "hash" before constructing the new filename and upload the full-text on S3. - Improve a log-message.Lampros Smyrnaios2023-03-29 17:12:37 +0300
495d5de19b- Automatically get the status of the docker containers after 30 secs of their initialization. - Add an error-handling in "installAndRun.sh" - Update dependencies.Lampros Smyrnaios2023-03-27 19:43:15 +0300
4280f89296- Set the default value of the "isTestEnvironment" property to "true", in order to avoid undesired outcomes in the production db. - Code polishing.Lampros Smyrnaios2023-03-21 17:04:28 +0200
e975bec911- Add Prometheus and Grafana which help measuring various metrics for the Controller's health and performance. - Fix Docker config still using the old (now removed) "application.properties" file. - Simplify the process of building and running the docker image; Now we use docker compose to run the Controller, along with the Prometheus and Grafana. Also, now it is not requested from the user to login and push the image (this may change in the future).Lampros Smyrnaios2023-03-21 16:46:33 +0200
003c0bf179- Add support for excluding specific datasources from being crawled. These datasources may be aggregated through bulk-imports, by other pieces of software. Such a datasource is "arXiv.org". - Fix an issue, where the "datasource-type" was retrieved instead of the "datasource-name". - Polish the "findAssignmentsQuery".Lampros Smyrnaios2023-03-21 07:19:35 +0200
f835a752bfTransform the "application.properties" file to "application.yml" and optimize the property-trees.Lampros Smyrnaios2023-03-20 15:23:00 +0200
4af298a52aRevert the version of "libthrift"-dependency to "0.17.0", as the newer version is not compatible with Java 8.Lampros Smyrnaios2023-03-03 12:57:30 +0200
7b217764e0Improve performance when downloading and decompressing the full-texts archive.Lampros Smyrnaios2023-03-02 17:44:53 +0200
8893662a81Refactor the UrlsController: a) offload the business-logic to the dedicated "UrlsService" and b) move the "checkParquetFilesSuccess()"-method to "ParquetFileUtils".Lampros Smyrnaios2023-02-21 15:36:35 +0200
a1c16ffc19- Exclude empty and null urls in the assignments. - Update the "getFullTextsImproved"-call to "getFullTexts", now that the "improved" version is stable. - Update Gradle. - Code polishing.Lampros Smyrnaios2023-02-16 14:24:47 +0200
2253f05bf5Refactor the "StatsController"-code, by offloading it to a dedicated "StatsService".Lampros Smyrnaios2023-02-09 19:25:48 +0200
49fefefafd- Refactor the payloads-statistics-code and provide two endpoints: "getNumberOfPayloadsAggregatedByService", which returns the number of payloads aggregated only by the PDF-Aggregation-Service, and the "getNumberOfAllPayloads", which returns the number of all payloads existing in the database, even the ones aggregated in the past, by other pieces of software. - Update README.md. - Make sure the docker image is clean-built, by avoiding the use of cache.Lampros Smyrnaios2023-02-02 17:58:47 +0200
c9f33d3afaAdd an extra precaution-check to allow the emptying or deletion of an S3-Object-Store bucket, only when the app runs in "TestEnvironment".Lampros Smyrnaios2023-02-01 16:42:22 +0200
dc8f0f2bd1- Heavily reduce the maximum amount of space needed, by deleting the files of each full-texts batch, right after they are uploaded to the S3 Object Store. - Add a check for when the retrieved full-texts-batch is missing some requested files and show a warn-log. - Update dependencies.Lampros Smyrnaios2023-01-23 20:23:21 +0200
d8773e6ebb- Make sure the test-environment uses a dedicated hdfs-parquet-directory. - Block app-execution in case the hdfs parquet directories failed to be created. - Code polishing.Lampros Smyrnaios2023-01-18 13:38:05 +0200
c08ba1cc89Revert the update of the "minio" dependency, as it introduces a bug, related to the "okhttp3.HttpUrl"-class.Lampros Smyrnaios2023-01-10 15:58:23 +0200
8876089022- Use Facebook's [**Zstandard**](https://facebook.github.io/zstd/) compression algorithm, which brings very big benefits on compression rate and speed. - Update the minIO dependency. - Code polishing.Lampros Smyrnaios2023-01-10 13:34:54 +0200
d1a4c84289- Make sure the fullPath of the baseFilesLocation is available when the user specifies a non-root directory. - Improve error-checking and exception-handling in some "S3ObjectStore"-methods. - Make sure the "responseCode" is "200-OK", before trying to get the InputStream in "UriBuilder.getPublicIP()".Lampros Smyrnaios2023-01-09 15:44:53 +0200
9904ea5743- Improve the stability of "UriBuilder.getPublicIP()", by using a "HttpURLConnection" to increase the connection and read timeouts and avoid timeout-exceptions. - Update Spring.Lampros Smyrnaios2023-01-03 18:39:50 +0200
4528d1f9be- Fix the "baseFilesLocation" being null (there was no serious problem, but multiple directories were spawned in the project's directory). - Make sure the given "baseFilesLocation" ends with a file-separator, before using it. - Optimize the process of unzipping-files.Lampros Smyrnaios2022-12-20 18:38:11 +0200
e11afe5ab2Improve performance of the hash-checking algorithm by using multithreading.Lampros Smyrnaios2022-12-15 18:34:28 +0200
e51ee9dd27- Add info about the Stats API usage in "README.md". - Optimize performance in "ParquetFileUtils.createAndLoadParquetDataIntoAttemptTable()" and "ParquetFileUtils.createAndLoadParquetDataIntoPayloadTable()". - Handle the "EmptyResultDataAccessException" inside "StatsController". - Optimize gradle's performance. - Code polishing.Lampros Smyrnaios2022-12-15 14:04:22 +0200
bfdf06bd09- Apply error-checking on individual CallableTasks and in tasks-batches related to the creation and upload of all the data related to the "attempt" and "payload" table. So, if no data could be uploaded for one or both tables, no "load"-queries will be executed for that/those tables. - Catch the more general "Exception", inside "FileUtils.mergeParquetFiles()", in order to be certain that the "SQLException" can also be caught. - Code polishing.Lampros Smyrnaios2022-12-09 12:46:06 +0200
0209d24068- Change the parquet compression from "Snappy" to "Gzip", as there is an unhandleable exception when the app is running inside a Docker Container and uses the "Snappy" compression. - Code polishing.Lampros Smyrnaios2022-12-08 16:28:41 +0200
c8baf5a5fc- Fix not finding the parquet-schema files when the app was run inside a Docker Container. - Update the "namespaces" and the "names" inside the parquet schemas. - Code polishing.Lampros Smyrnaios2022-12-08 12:16:05 +0200
95c38c4a24- Fix creating the "assignment" table, always in the testDatabase. - Code polishing.Lampros Smyrnaios2022-12-07 14:58:38 +0200
f183df276b- Move the "uploadFullTexts"-code in its own method. - Code polishing.Lampros Smyrnaios2022-12-06 12:24:34 +0200
b0c57d79a5- When the Controller cannot retrieve any assignments from Impala (without an error), return an HTTP-"MULTI_STATUS" with an empty "AssignmentsResponse", instead of an "INTERNAL_SERVER_ERROR". - Fix an error-message.Lampros Smyrnaios2022-12-05 16:44:00 +0200
577ea983e8- Improve some log-messages. - Set some optimization settings for gradle. - Fix error-handling in "installAndRun.sh". - Update dependencies.Lampros Smyrnaios2022-11-30 16:28:39 +0200
6226e2298d- Upgrade the results-loading process: Instead of making thousands of sql-insert requests to Impala now we write the results to parquet files, upload them to HDFS and then import the data into the Impala tables with just 2 requests. This results in a huge performance improvement. One side effect of using the parquet-files, is that the timestamps are now BIGDECIMAL numbers, instead of "Timestamp" objects, but, converting them to such objects is pretty easy, if we ever need to do it. - Code polishing.Lampros Smyrnaios2022-11-10 17:18:21 +0200
Lampros Smyrnaios