add robustness to avoid crashing on unexepected input

[colinsurprenant]

this is a followup from issue #1240 - where pre multiline/tv_sec fix shippers were paired with post multiline/tv_sec fix indexers using redis. The indexer was crashing on the tv_sec error because the shipper malformed the event.

should we add robustness to avoid crashing in such a condition?

[colinsurprenant]

also mentioned in https://logstash.jira.com/browse/LOGSTASH-1623

[colinsurprenant]

same idea in https://logstash.jira.com/browse/LOGSTASH-2107

[KlavsKlavsen]

I haven't looked into how exceptions are handled in logstash currently, so I'm just spitballing here :)

My feeling is, that handling exceptions needs to somehow be centralized - so it can be handled better once - and for all times. I'm sure you are better than me, at designing that in java :)

some sort of automatic exception handling that each input, output, filter etc. simply inherits (and can extend if need be), which does something default (and configurable) - like drop the line being worked on to a "failed lines" output - and then just skip and continue.

[suyograo]

related to #2130 (comment)

[colinsurprenant]

the global exception handling issue is #2386 - I will keep this one open to make sure this specific concerned is tracked.

[henriklynggaard]

Any traction on this. It has been open for a long time?

[colinsurprenant]

I am closing this issue and redirecting to #4127. A lot a progress has been made to improve exception handling & reliability and the original issue reported here is not very current anymore.