Just read this thread:
https://issues.apache.org/jira/browse/SPARK-6305

Concerns: 
* User-supplied logging configurations would break 
* Spark code to suppress some overly verbose logging in dependencies (which 
ignore config?) -not sure about impact 
* POM changes to exclude log4j 1.2 - Spark has many many dependencies which 
bring it back in 

Mikael, let me know if I missed something. 

We can do something about #1. 
Not sure about #2 and #3. 

Remko


Sent from my iPhone

> On 2016/08/13, at 7:38, Matt Sicker <[email protected]> wrote:
> 
> Yeah, really, what's the alternative?
> 
> And I do think that completing that Log4j1ConfigurationFactory would be a 
> great way to get people to migrate if they don't even need to update configs. 
> Though I'd hate to come across one of those configs in the wild because it's 
> not nearly as well documented as the 2.x style configs are.
> 
>> On 12 August 2016 at 15:45, Ralph Goers <[email protected]> wrote:
>> Dropping Log4j 1 and doing what instead? Stop logging altogether?  Logback’s 
>> configuration isn’t compatible either.
>> 
>> Ralph
>> 
>>> On Aug 12, 2016, at 9:52 AM, Mikael Ståldal <[email protected]> 
>>> wrote:
>>> 
>>> There seem to be quite some frustration in some other Apache projects about 
>>> migrating from Log4j 1. Especially around configration. Some seem to 
>>> consider dropping Log4j completely instead of going for Log4j 2.
>>> 
>>> I have taken on Spark, Hadoop and Zookeeper; see links to their Jira issues 
>>> from LOG4J2-1473. I could need some help.
>>> 
> 
> 
> 
> -- 
> Matt Sicker <[email protected]>

Reply via email to