Hi ShaoFeng, I'll create an issue in JIRA later. Thanks for your reply.
Regards, Liang On Tue, Dec 10, 2019 at 6:17 PM ShaoFeng Shi <[email protected]> wrote: > Yes, Kylin supports in parallel segment building. It seems to be a bug > introduced in KYLIN-3977 in Kylin 2.6.3. Would you like to create it as a > JIRA so that you can get notified when there is a hot-fix? The JIRA can be > reported in https://issues.apache.org/jira/secure/Dashboard.jspa , and > selecting KYLIN as the project. Thank you! > > > Best regards, > > Shaofeng Shi 史少锋 > Apache Kylin PMC > Email: [email protected] > > Apache Kylin FAQ: https://kylin.apache.org/docs/gettingstarted/faq.html > Join Kylin user mail group: [email protected] > Join Kylin dev mail group: [email protected] > > > > > liang <[email protected]> 于2019年12月10日周二 下午5:43写道： > >> Hi there, >> We noticed a segment refresh job at ERROR status at "*2019-12-10 >> 02:27:15*". After a deep dive into kylin.log, we found that the >> WriteConflictException was raised when trying to update the last modified >> time for the dict. >> >> (Sensitive content are consered) >> >> 2019-12-10 02:27:15,576 ERROR [Scheduler 1186524190 Job >>> f76fb12e-477f-b404-e021-adaf520d04bb-321] common.HadoopShellExecutable:65 : >>> error execute >>> HadoopShellExecutable{id=f76fb12e-477f-b404-e021-adaf520d04bb-03, >>> name=Build Dimension Dictionary, state=RUNNING} >>> org.apache.kylin.common.persistence.WriteConflictException: Overwriting >>> conflict >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict, >>> expect old TS 1575916011421, but it is 1575916035573 >>> at >>> org.apache.kylin.storage.hbase.HBaseResourceStore.updateTimestampImpl(HBaseResourceStore.java:372) >>> at >>> org.apache.kylin.common.persistence.ResourceStore.lambda$updateTimestampWithRetry$4(ResourceStore.java:443) >>> at >>> org.apache.kylin.common.persistence.ExponentialBackoffRetry.doWithRetry(ExponentialBackoffRetry.java:52) >>> at >>> org.apache.kylin.common.persistence.ResourceStore.updateTimestampWithRetry(ResourceStore.java:442) >>> at >>> org.apache.kylin.common.persistence.ResourceStore.updateTimestampCheckPoint(ResourceStore.java:437) >>> at >>> org.apache.kylin.common.persistence.ResourceStore.updateTimestamp(ResourceStore.java:432) >>> at >>> org.apache.kylin.dict.DictionaryManager.updateExistingDictLastModifiedTime(DictionaryManager.java:197) >>> at >>> org.apache.kylin.dict.DictionaryManager.trySaveNewDict(DictionaryManager.java:157) >>> at >>> org.apache.kylin.dict.DictionaryManager.saveDictionary(DictionaryManager.java:339) >>> at >>> org.apache.kylin.cube.CubeManager$DictionaryAssist.saveDictionary(CubeManager.java:1145) >>> at >>> org.apache.kylin.cube.CubeManager.saveDictionary(CubeManager.java:1107) >>> at >>> org.apache.kylin.cube.cli.DictionaryGeneratorCLI.processSegment(DictionaryGeneratorCLI.java:100) >>> at >>> org.apache.kylin.cube.cli.DictionaryGeneratorCLI.processSegment(DictionaryGeneratorCLI.java:69) >>> at >>> org.apache.kylin.engine.mr.steps.CreateDictionaryJob.run(CreateDictionaryJob.java:73) >>> at org.apache.kylin.engine.mr.MRUtil.runMRJob(MRUtil.java:93) >>> at >>> org.apache.kylin.engine.mr.common.HadoopShellExecutable.doWork(HadoopShellExecutable.java:63) >>> at >>> org.apache.kylin.job.execution.AbstractExecutable.execute(AbstractExecutable.java:167) >>> at >>> org.apache.kylin.job.execution.DefaultChainedExecutable.doWork(DefaultChainedExecutable.java:71) >>> at >>> org.apache.kylin.job.execution.AbstractExecutable.execute(AbstractExecutable.java:167) >>> at >>> org.apache.kylin.job.impl.threadpool.DefaultScheduler$JobRunner.run(DefaultScheduler.java:114) >>> at >>> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) >>> at >>> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) >>> at java.lang.Thread.run(Thread.java:745) >>> >> >> We have a glance at the codebase (Kylin-2.6.4), we found the last >> modified time of the dict (which has been built) will be updated at the >> step "Build Dimension Dictionary". The log blow shows the same dict is >> processed many times. >> >> Liangs-MacBook-Pro:~ pwrliang$ cat 12-10.log |grep "has already been >>> built, save it"|grep "THIS_IS_COLUMN_NAME" >>> 2019-12-10 02:07:10,103 DEBUG [Scheduler 1186524190 Job >>> 5c4b2ee9-c8aa-9a69-ffb0-3a6811ad22a9-372] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:07:11,036 DEBUG [Scheduler 1186524190 Job >>> 593979b8-e80b-590f-335b-526c2fc49080-295] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:07:20,455 DEBUG [Scheduler 1186524190 Job >>> 160952f3-fef2-4fdf-c08a-77d49f830805-246] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:07:20,888 DEBUG [Scheduler 1186524190 Job >>> 69304c1c-cc02-3bbe-67d4-efdd3858e88a-234] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:07:21,634 DEBUG [Scheduler 1186524190 Job >>> 345b688c-7992-f2cf-ae06-835675f545ea-117] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:07:23,040 DEBUG [Scheduler 1186524190 Job >>> 0ea711ed-71b5-ae47-a34a-f0a27fd9d7cc-303] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:07:36,358 DEBUG [Scheduler 1186524190 Job >>> 6daf351f-5073-128f-8754-f499e345db54-308] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:26:43,716 DEBUG [Scheduler 1186524190 Job >>> c90a765c-976f-49c2-38d8-e6812d091cfb-295] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:26:44,944 DEBUG [Scheduler 1186524190 Job >>> cd108004-12c6-078f-8646-f29841de9e40-205] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:26:51,306 DEBUG [Scheduler 1186524190 Job >>> d808dadf-8eea-34db-1b1f-d50e923b4e3a-372] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:27:15,523 DEBUG [Scheduler 1186524190 Job >>> ea07ef8f-b2b7-b3f8-c481-df615025217d-117] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:27:15,535 DEBUG [Scheduler 1186524190 Job >>> f76fb12e-477f-b404-e021-adaf520d04bb-321] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> 2019-12-10 02:29:42,504 DEBUG [Scheduler 1186524190 Job >>> f894f427-b170-19b6-76fb-e1b671659771-321] cli.DictionaryGeneratorCLI:99 : >>> Dict for 'THIS_IS_COLUMN_NAME' has already been built, save it >>> >> >> Since we issue the refresh segments request concurrently (with REST >> APIs), the dict has a chance to be modified concurrently. Kylin uses >> checkAndPut >> to update HBase to prevent race-condition, so the WriteConflictException >> may be raised when updating the same resource. >> >> Liangs-MacBook-Pro:~ pwrliang$ cat 12-10.log |grep >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict|grep >>> "Update dictionary " >>> 2019-12-10 02:07:10,223 INFO [Scheduler 1186524190 Job >>> 5c4b2ee9-c8aa-9a69-ffb0-3a6811ad22a9-372] dict.DictionaryManager:198 : >>> Update dictionary >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict >>> lastModifiedTime to 1575914830203 >>> 2019-12-10 02:07:11,109 INFO [Scheduler 1186524190 Job >>> 593979b8-e80b-590f-335b-526c2fc49080-295] dict.DictionaryManager:198 : >>> Update dictionary >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict >>> lastModifiedTime to 1575914831090 >>> 2019-12-10 02:07:20,554 INFO [Scheduler 1186524190 Job >>> 160952f3-fef2-4fdf-c08a-77d49f830805-246] dict.DictionaryManager:198 : >>> Update dictionary >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict >>> lastModifiedTime to 1575914840528 >>> 2019-12-10 02:07:21,762 INFO [Scheduler 1186524190 Job >>> 345b688c-7992-f2cf-ae06-835675f545ea-117] dict.DictionaryManager:198 : >>> Update dictionary >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict >>> lastModifiedTime to 1575914841727 >>> 2019-12-10 02:07:23,253 INFO [Scheduler 1186524190 Job >>> 0ea711ed-71b5-ae47-a34a-f0a27fd9d7cc-303] dict.DictionaryManager:198 : >>> Update dictionary >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict >>> lastModifiedTime to 1575914843095 >>> 2019-12-10 02:26:43,843 INFO [Scheduler 1186524190 Job >>> c90a765c-976f-49c2-38d8-e6812d091cfb-295] dict.DictionaryManager:198 : >>> Update dictionary >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict >>> lastModifiedTime to 1575916003771 >>> 2019-12-10 02:26:51,436 INFO [Scheduler 1186524190 Job >>> d808dadf-8eea-34db-1b1f-d50e923b4e3a-372] dict.DictionaryManager:198 : >>> Update dictionary >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict >>> lastModifiedTime to 1575916011421 >>> 2019-12-10 02:27:15,593 INFO [Scheduler 1186524190 Job >>> ea07ef8f-b2b7-b3f8-c481-df615025217d-117] dict.DictionaryManager:198 : >>> Update dictionary >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict >>> lastModifiedTime to 1575916035573 >>> 2019-12-10 02:29:42,564 INFO [Scheduler 1186524190 Job >>> f894f427-b170-19b6-76fb-e1b671659771-321] dict.DictionaryManager:198 : >>> Update dictionary >>> /dict/THIS_IS_CUBE_NAME/THIS_IS_COLUMN_NAME/010dbf72-52a9-e759-08a7-ed7cde0c6e0d.dict >>> lastModifiedTime to 1575916182549 >>> >> >> The log above shown that the job "f76fb12e-477f-b404-e021-adaf520d04bb" >> has not been print out the log - "Update dictionary blablabla...." >> because of the WriteConflictException. >> >> The question is: refreshing segments concurrently is supported for Kylin >> 2.6.4? Concurrently refreshing is very important for us, since we want to >> reduce the building time. >> >> Regards, >> Liang >> >
