yesterday we tried to update our production system.
we have N logstash instances (as shipper) to a few redis instances.
Then there are a few logstash indexer pulling from redis and putting them into the elasticsearch cluster.
with logstash 1.5.1 and elasticsearch 1.6.1 all was fines for weeks.
Now, with logstash 1.5.3 (JDK 7 and 8) we run into the problem that logstash is dying with notification like
"undefined method `close' for nil:NilClass".
Since i am no ruby expert, can anybody be of assistance what went wrong and how to fix this ?
Elasticsearch runs but does not receive events from logstash since logstash dies after nearly 30 sec.
I checked our plugins, they were ok so far i could see.
Meanwhile we found the error causing this:
In the configuration folder, where logstash looks for configuration files never ever has to be a folder!.
If, say /tmp/conf is your config folder, then in this folder only files should be located.
Otherwise logstash dies with NO error output useful to find this bug.
Is it worth an issue on github? or who can check this?
you call your logstash with "-f <path_to_config_file_or_folder>".
In our case the path was a folder like e.g. "logstash -f /tmp/ls_configs" where our logstash configuration resides.
The problem then was that inside the folder "/tmp/ls_configs" there was another folder e.g. "/tmp/ls_configs/foo/".
This caused the problem since logstash couldn't handle subfolders in the config path.
After we removed the "foo" folder from the config folder the problem was gone.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.