I believe this is handled by the Logviewer daemon. If you run "storm logviewer" on your worker machines, the Logviewer daemon will periodically delete old topology directories.
Take a look at https://github.com/apache/storm/blob/9680928a0c6e26a8a2d3e5ceb9a632579546353d/storm-webapp/src/main/java/org/apache/storm/daemon/logviewer/utils/LogCleaner.java to see how this works. Den tor. 4. apr. 2019 kl. 10.53 skrev Mohit Goyal <itsmohitgo...@gmail.com>: > Hi, > > I'm using log4j to log all my topology logs in apache storm. Using the > following standard format I'm generating appender files: > > >> fileName="${sys:workers.artifacts}/${sys:storm.id >> }/${sys:worker.port}/${sys:logfile.name}" >> filePattern="${sys:workers.artifacts}/${sys:storm.id >> }/${sys:worker.port}/${sys:logfile.name}.%i"> > > Now the problem I'm facing here is that if I deploy any topology again, > it'll create a separate folder and create log files in that folder since the > new topology will have different value of *sys:storm.id <http://storm.id/>* > field. This way my supervisor machine will still have logs from older version > of topology, which over the time accumulates and cause me some memory issues. > Has anyone hare faced this issue and what can be the possible solution. One > solution can be that I remove *sys:storm.id <http://storm.id/>* field from > filename but since I've several topologies running, I don't want to remove > topology name from my log file path. > > > Thanks and regards, > Mohit Goyal > >
