Denny Ye created FLUME-1644:
-------------------------------
Summary: Shouldn't create 'log-xxx.log' file infinitely at space
limitation local environment
Key: FLUME-1644
URL: https://issues.apache.org/jira/browse/FLUME-1644
Project: Flume
Issue Type: Bug
Components: Channel
Affects Versions: v1.2.0
Reporter: Denny Ye
Fix For: v1.3.0
There have consuming failure at FileChannel and Flume didn't delete
inactive/consumed log file any more. Thus, 'No space left on device' was
happening.
I want to raise another problem that in this case, there have too many log file
with zero length. Flume holds so many file descriptors. Unluckily, Flume also
creates log file every fixed minutes.
Look at my storage folder in local of Flume.
{code:xml}
-rw------- 1 root root 0 Oct 16 00:26 log-6468.meta.tmp
-rw------- 1 root root 0 Oct 16 00:26 log-6469
-rw------- 1 root root 0 Oct 16 00:26 log-6469.meta.tmp
-rw------- 1 root root 0 Oct 16 00:26 log-6464
-rw------- 1 root root 0 Oct 16 00:26 log-6464.meta.tmp
-rw------- 1 root root 0 Oct 16 00:26 log-6465
-rw------- 1 root root 0 Oct 16 00:26 log-6465.meta.tmp
-rw------- 1 root root 0 Oct 16 00:26 log-6462
-rw------- 1 root root 0 Oct 16 00:26 log-6462.meta.tmp
-rw------- 1 root root 0 Oct 16 00:26 log-6463
-rw------- 1 root root 0 Oct 16 00:26 log-6463.meta.tmp
-rw------- 1 root root 0 Oct 16 00:26 log-6460
-rw------- 1 root root 0 Oct 16 00:26 log-6460.meta.tmp
-rw------- 1 root root 0 Oct 16 00:26 log-6461
-rw------- 1 root root 0 Oct 16 00:26 log-6461.meta.tmp
-rw------- 1 root root 0 Oct 16 00:25 log-6458
-rw------- 1 root root 0 Oct 16 00:25 log-6458.meta.tmp
-rw------- 1 root root 0 Oct 16 00:25 log-6459
-rw------- 1 root root 0 Oct 16 00:25 log-6459.meta.tmp
-rw------- 1 root root 0 Oct 16 00:25 log-6456
-rw------- 1 root root 0 Oct 16 00:25 log-6456.meta.tmp
-rw------- 1 root root 0 Oct 16 00:25 log-6457
-rw------- 1 root root 0 Oct 16 00:25 log-6457.meta.tmp
-rw------- 1 root root 0 Oct 16 00:25 log-6454.meta.tmp
-rw------- 1 root root 0 Oct 16 00:25 log-6455
...
...
{code}
It might be retry by SinkRunner when it will rollback in time, and invoke
'roll' method at every failure cases.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
