flume-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "taoyang (Jira)" <j...@apache.org>
Subject [jira] [Created] (FLUME-3352) Unnecessary canary test will block on readonly spooldir while another trackerdir is set.
Date Thu, 16 Jan 2020 11:57:00 GMT
taoyang created FLUME-3352:
------------------------------

             Summary: Unnecessary canary test will block on readonly spooldir while another
trackerdir is set.
                 Key: FLUME-3352
                 URL: https://issues.apache.org/jira/browse/FLUME-3352
             Project: Flume
          Issue Type: Bug
          Components: Sinks+Sources
    Affects Versions: 1.9.0
            Reporter: taoyang
             Fix For: 1.10.0


*Phenomenon*

In many cases, we have just read permission on spoolDir and write permission on trackerDir.

However whenever flume starts for spooldir source , it will always try to create a '.canary'
file in the spooling directory.

Then it leads to the failure of some processing unnecessarily.

*Recur*

(Usually spooldir is mounted readonly from a nas in production and this time we create a readonly
directory instead.)

First we create the spoolDir by root user and create a file in it and it is readonly for others
by default.
{code:java}
su root
mkdir /home/hadoop/testspooldir
echo 'foo' > /home/hadoop/testspooldir/bar{code}
Then switch to another user (hadoop) who runs flume and make sure it has read permission for
the spooldir.
{code:java}
su hadoop
mkdir /home/hadoop/testtrackerdir
ll /home/hadoop/testspooldir
>> total 4-rw-r--r-- 1 root root 4 Jan 16 19:15 bar{code}
now create the example.conf:
{code:java}
a1.sources = r1
a1.sinks = k1
a1.channels = c1
a1.sources.r1.type = spooldir
a1.sources.r1.deletePolicy = never
a1.sources.r1.spoolDir = /home/hadoop/testspooldir
a1.sources.r1.trackerDir = /home/hadoop/testtrackerdir
a1.sources.r1.trackingPolicy = tracker_dir
a1.sinks.k1.type = logger
a1.channels.c1.type = memory
a1.channels.c1.capacity = 1000
a1.channels.c1.transactionCapacity = 100
a1.sources.r1.channels = c1
a1.sinks.k1.channel = c1
{code}
and start flume with it
{code:java}
bin/flume-ng agent --conf conf --conf-file conf/example.conf --name a1 -Dflume.root.logger=INFO,console{code}
then the IOException is thrown.
{code:java}
2020-01-16 19:16:12,777 (lifecycleSupervisor-1-0) [ERROR - org.apache.flume.lifecycle.LifecycleSupervisor$MonitorRunnable.run(LifecycleSupervisor.java:251)]
Unabl2020-01-16 19:16:12,777 (lifecycleSupervisor-1-0) [ERROR - org.apache.flume.lifecycle.LifecycleSupervisor$MonitorRunnable.run(LifecycleSupervisor.java:251)]
Unable to start EventDrivenSourceRunner: { source:Spool Directory source r1: { spoolDir: /home/hadoop/testspooldir
} } - Exception follows.org.apache.flume.FlumeException: Unable to read and modify files in
the spooling directory: /home/hadoop/testspooldir at org.apache.flume.client.avro.ReliableSpoolingFileEventReader.<init>(ReliableSpoolingFileEventReader.java:195)
at org.apache.flume.client.avro.ReliableSpoolingFileEventReader.<init>(ReliableSpoolingFileEventReader.java:89)
at org.apache.flume.client.avro.ReliableSpoolingFileEventReader$Builder.build(ReliableSpoolingFileEventReader.java:882)
at org.apache.flume.source.SpoolDirectorySource.start(SpoolDirectorySource.java:111) at org.apache.flume.source.EventDrivenSourceRunner.start(EventDrivenSourceRunner.java:44)
at org.apache.flume.lifecycle.LifecycleSupervisor$MonitorRunnable.run(LifecycleSupervisor.java:249)
at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) at java.util.concurrent.FutureTask.runAndReset(FutureTask.java:308)
at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.access$301(ScheduledThreadPoolExecutor.java:180)
at java.util.concurrent.ScheduledThreadPoolExecutor$ScheduledFutureTask.run(ScheduledThreadPoolExecutor.java:294)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
at java.lang.Thread.run(Thread.java:748)Caused by: java.io.IOException: Permission denied
at java.io.UnixFileSystem.createFileExclusively(Native Method) at java.io.File.createTempFile(File.java:2024)
at org.apache.flume.client.avro.ReliableSpoolingFileEventReader.<init>(ReliableSpoolingFileEventReader.java:185)
... 12 more{code}
*Fix*

We just add the condition where this trick is necessary.

The pr/patch will be submitted as as shown below.

Or let it still exist and using warning log instead of exception thrown is better?



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@flume.apache.org
For additional commands, e-mail: issues-help@flume.apache.org


Mime
View raw message