Is this library ready to use in production like environment ?

[mayurjurani]

Hi @urielha / @jptoto

I was really wondering as there is no documentation available for this project that, is this library ready to use in production ?

[urielha]

Hi,
Sure - the library is already in use in several projects for about 3 years.
All of the known issues have already been solved and we trying to do our best to give answers if any new issue is found.

I wonder, what kind of documentation were you looking for?
At the main project's README you have examples of configurations and how to use which seems to be enough for a small scale project like this.

[jptoto]

I can't speak for this fork but I would not use my lib for production as it's still based on blocking log4net calls internally. If Elasticsearch becomes unavailable you will run into performance issues.

[eran-gil]

This has been an issue with log4stash as well, but it was fixed more than a year ago.
log4stash supports async logging, which means it doesn't wait for a response from ElasticSearch, and therefore is not affected from unavailability of your ElasticSearch cluster.
Log4stash also added support for multi-server configuration, which reduces the load on your cluster's nodes even more.
Hope this helps @mayurjurani .

[mayurjurani]

Thank you so much for your inputs. @urielha @jptoto @erangil2 This is extremely helpful.

[mayurjurani]

@urielha / @erangil2 ,

One information I am looking is, what will happen to those async messages (threads) if ElasticSearch cluster goes offline? Will those messages/threads get killed or they will be persisted in memory until ElasticSearch cluster becomes available ?
Our application produces lots of useful logs that we need to push to our ELK cluster.

Did you guys performed these tests ? if so, could you please share the test results ?
Also, can you please recommend what sort of testing I should do in-order to test such behavior ?

[eran-gil]

If the cluster is down, there is a timeout configuration for the appender, and if the timeout passes, the thread is killed and the message is lost.
In my opinion an application should be resistent to log loss unlike data which should be persisted until is written.