Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.
Sign upDeprecate multiline filter plugin in favor of multiline codec #4386
Comments
suyograo
added
release plan
v2.2.0
labels
Dec 22, 2015
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
runningman84
Jan 7, 2016
Contributor
Please do the multiline handling at filebeat. Right now filebeat might send a events to two different logstash servers which would prevent it from beeing processed as a mutline event.
|
Please do the multiline handling at filebeat. Right now filebeat might send a events to two different logstash servers which would prevent it from beeing processed as a mutline event. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
|
@runningman84 absolutely. elastic/filebeat#89 |
suyograo
closed this
Jan 11, 2016
ChrisMagnuson
referenced this issue
Apr 5, 2016
Closed
Multiline: message stays in buffer until new log entry #1482
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
mrjameshamilton
Oct 12, 2016
How can I have different multiline patterns for different types of files? Currently, I have defined several multiline patterns and use ifs to choose between them based on the filename. Is this possible with the multiline codec? It seems that I can only define a single pattern.
mrjameshamilton
commented
Oct 12, 2016
|
How can I have different multiline patterns for different types of files? Currently, I have defined several multiline patterns and use ifs to choose between them based on the filename. Is this possible with the multiline codec? It seems that I can only define a single pattern. |
This comment has been minimized.
Show comment
Hide comment
This comment has been minimized.
rahulghanate
commented
Oct 25, 2016
|
I have the similar issue what @mrjameshamilton have mentioned. |
suyograo commentedDec 22, 2015
Multiline (ML) event processing is a popular use case in Logstash. Multiline event processing is complex and relies on proper event ordering. The best way to guarantee ordered log processing is to implement the processing as early in the pipeline as possible.
Today, users have 2 choices - multiline filter and multiline codec. The preferred tool in the Logstash pipeline is the multiline codec, which merges lines from a single input using a simple set of rules. This can be used with any input source.
Multiline filter performs a similar task of aggregating multiple lines, and exists because it predates the
codecconcept in Logstash. This filter is not thread-safe and was advised not to be used when settingfilter_workeror-wto more than 1.Recently, there has been a number of key enhancements in ML codec - proper handling of streams from multiple sources, auto-flushing of buffered lines, circuit breakers for preventing OOMs and so on. At this point we feel ML filter has run its course, and needs to be deprecated. This would also avoid confusion for new users - having to pick between the two.
The plan is to deprecate ML filter in version 2.2, and remove it from the default package in version 3.0. You can still install ML filter by using
bin/plugin install, but our advice is to move to the more robust ML codec.