accumulo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Christopher <>
Subject Re: Getting started with Accumulo-1070 (improve audit messages)
Date Wed, 01 May 2013 11:48:26 GMT
Responses inline below:

On Wed, May 1, 2013 at 4:30 AM, Rob Tallis <> wrote:
> Hi
> I'm new to accumulo, new to open source projects. I want to pick up
> Accumulo-1070 (Improve Audit Messages) from where Phil Young left off. I've
> a bunch of questions about how I go about this.
>    - Why is there an svn and a github repository? Which should I be using?

SVN is our primary. GIT is a mirror and may lag commits to SVN. You
can use whichever is more convenient for you.

>    - What's the workflow for me submitting a change? Do I upload a patch
>    file to the ticket? We then go back and forward on the ticket until you're
>    happy with the patch?

Yes, essentially, you create a patch and attach it to a ticket to contribute.

> On the specific task, here's what I'd like to do:
>    - In  - Phil and Keith found that without
>    a getInstance() implementation here, the bulk of that class is never used
>    and the audit messages already in there haven't seen the light of day. Once
>    fixed, I want to rewrite the messages so they are less "prose" and more
>    machine-readable.
>    - Get rid of the custom and use a separate Audit Logger.
>    - Stick the Logger and Appender config in a new log4j xml config file.
>    Sites can configure easily where they want the messages to go. I want to
>    stick all the audit messages in a separate file (or files).
>    - (maybe) implement config to enable/disable auditing based on user.

Sounds reasonable to me. Your patch may get reviewed by one or more
committers before acceptance, though.

> In the ticket comments Keith said:
>    - "Will be logging binary data,this will be lossy. For example it will
>    log the columns a user fetches. Could consider encoding base64 non-ascii
>    characters, but I am not sure about this."
> I don't understand this, column names can be binary? Where does that happen?

Yes, essentially fields in Accumulo (except the timestamp and
visibility) are arbitrary bytes. They may or may not represent
human-readable strings. We don't enforce a constraint that they be
human-readable strings.

> Also, do you have an indication when 1.5 will get released? I need to do a
> local patch for the "current" version to keep my org happy, if 1.5 is
> imminent I won't need to bother with 1.4 (this is just for my use).

I'm planning to roll out a release candidate Friday for a test vote.

> Thanks,
> Rob Tallis

Christopher L Tubbs II

View raw message