Thanks Hari, I made that change, will see what happens.
Regards, Guy Needham | Data Discovery Virgin Media | Technology and Transformation | Data Bartley Wood Business Park, Hook, Hampshire RG27 9UP D 01256 75 3362 I welcome VSRE emails. Learn more at http://vsre.info/ ________________________________ From: Hari Shreedharan [mailto:hshreedha...@cloudera.com] Sent: 13 January 2015 18:06 To: user@flume.apache.org Cc: user@flume.apache.org Subject: Re: Many files in file channel data directory This issue is that your max file size is really small. This parameter: flume_test.channels.ch1.maxFileSize = 10000000 sets the size of each file in bytes, so you want it to be much bigger. Why not simply leave it at default which will keep at around 1.6G? These are not your HDFS files' sizes, these are the file sizes for internal file sizes of the file channel. Thanks, Hari On Tue, Jan 13, 2015 at 7:32 AM, Needham, Guy <guy.need...@virginmedia.co.uk<mailto:guy.need...@virginmedia.co.uk>> wrote: I'm running Flume 1.5.0 with this configuration: flume_test.sources = sr1 flume_test.channels = ch1 flume_test.sinks = sk1 #avro source flume_test.sources.sr1.type = avro flume_test.sources.sr1.channels = ch1 flume_test.sources.sr1.bind = 10.92.211.22 flume_test.sources.sr1.port = 55000 flume_test.sources.sr1.ssl = true flume_test.sources.sr1.keystore = /nas/used_by_hadoop/hadoop-kn-p2/rdd/hadoop_keystore.jks flume_test.sources.sr1.keystore-password = ***** flume_test.sources.sr1.compression-type = gzip #custom interceptor flume_test.sources.sr1.interceptors = i1 flume_test.sources.sr1.interceptors.i1.type = com.vm.rdd.TimeBodyInterceptor$Builder #memory channel flume_test.channels.ch1.type = file flume_test.channels.ch1.checkpointDir = /hadoop/user/flume/channels/flumeTest/checkpoint flume_test.channels.ch1.dataDirs = /hadoop/user/flume/channels/flumeTest/data flume_test.channels.ch1.capacity = 100000000 flume_test.channels.ch1.transactionCapacity = 10000 flume_test.channels.ch1.maxFileSize = 10000000 #HDFS channel flume_test.sinks.sk1.channel = ch1 flume_test.sinks.sk1.type = hdfs #dynamic path flume_test.sinks.sk1.hdfs.path = hdfs:///landing/data/flumeTest/%Y-%m-%d flume_test.sinks.sk1.hdfs.inUsePrefix = _ flume_test.sinks.sk1.hdfs.codeC = gzip #write at this block size when reach it flume_test.sinks.sk1.hdfs.rollSize = 2560000000 # roll file every 2 minutes if not filled a block flume_test.sinks.sk1.hdfs.rollInterval = 120 # if file has been idle for 15s, close file flume_test.sinks.sk1.hdfs.idleTimeout = 15 flume_test.sinks.sk1.hdfs.rollCount = 0 flume_test.sinks.sk1.hdfs.batchSize = 100 When I look at the data dir, there are many files: -rw-r--r-- 1 rdd rdd 0 Jan 13 12:06 in_use.lock -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-1 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-2 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-3 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-4 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-5 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-6 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-7 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-8 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-9 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-10 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-11 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-12 -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-13.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-1.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-2.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-3.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-4.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-5.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-6.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-7.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-8.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-9.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-10.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-11.meta -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-12.meta -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-13 -rw-r--r-- 1 rdd rdd 1.0M Jan 13 12:09 log-14 -rw-r--r-- 1 rdd rdd 47 Jan 13 12:09 log-14.meta -rw-r--r-- 1 rdd rdd 0 Jan 13 12:15 log-15 -rw-r--r-- 1 rdd rdd 47 Jan 13 12:15 log-15.meta This list can grow up to hundreds of log files which are all accessed around the same time. When looking in the log file, it appears that the channels are duplicates of one another, as the agent is writing to each channel at the same time, for example: 2015-01-13 12:09:52,451 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.Log.writeCheckpoint(Log.java:1005)] Updated checkpoint for file: /hadoop/user/flume/channels/flumeTest/data/log-14 position: 120431 logWriteOrderID: 1421150977639 2015-01-13 12:09:52,451 (Log-BackgroundWorker-ch1) [DEBUG - org.apache.flume.channel.file.Log.removeOldLogs(Log.java:1067)] Files currently in use: [14] 2015-01-13 12:09:52,451 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-1 2015-01-13 12:09:52,456 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-2 2015-01-13 12:09:52,461 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-3 2015-01-13 12:09:52,467 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-4 2015-01-13 12:09:52,472 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-5 2015-01-13 12:09:52,477 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-6 2015-01-13 12:09:52,482 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-7 2015-01-13 12:09:52,487 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-8 2015-01-13 12:09:52,492 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-9 2015-01-13 12:09:52,497 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-10 2015-01-13 12:09:52,503 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-11 2015-01-13 12:09:52,508 (Log-BackgroundWorker-ch1) [INFO - org.apache.flume.channel.file.LogFile$RandomReader.close(LogFile.java:504)] Closing RandomReader /hadoop/user/flume/channels/flumeTest/data/log-12 Does anyone know why there are so many active files at one time? Is this expected behaviour? Regards, Guy Needham | Data Discovery Virgin Media | Technology and Transformation | Data Bartley Wood Business Park, Hook, Hampshire RG27 9UP D 01256 75 3362 I welcome VSRE emails. Learn more at http://vsre.info/ -------------------------------------------------------------------- Save Paper - Do you really need to print this e-mail? Visit www.virginmedia.com for more information, and more fun. This email and any attachments are or may be confidential and legally privileged and are sent solely for the attention of the addressee(s). If you have received this email in error, please delete it from your system: its use, disclosure or copying is unauthorised. Statements and opinions expressed in this email may not represent those of Virgin Media. Any representations or commitments in this email are subject to contract. Registered office: Media House, Bartley Wood Business Park, Hook, Hampshire, RG27 9UP Registered in England and Wales with number 2591237 -------------------------------------------------------------------- Save Paper - Do you really need to print this e-mail? Visit www.virginmedia.com for more information, and more fun. This email and any attachments are or may be confidential and legally privileged and are sent solely for the attention of the addressee(s). If you have received this email in error, please delete it from your system: its use, disclosure or copying is unauthorised. Statements and opinions expressed in this email may not represent those of Virgin Media. Any representations or commitments in this email are subject to contract. Registered office: Media House, Bartley Wood Business Park, Hook, Hampshire, RG27 9UP Registered in England and Wales with number 2591237
