On Jun 14, 2013 8:34 PM, "Matthew Miller" mattdm@fedoraproject.org wrote:

Sooooo. As of systemd 198-1, (March 7th), systemd enables persistent

logging

of the journal by default -- that means, it's writing it to disk.

I'm not really excited about logging every event twice in the cloud image. The journal will do handwavy magic to automatically keep from running out

of

space, but until it hits its threshold, it _will_ be using up more disk.

I posted a message to the main devel list last year sometime about what realistic requirements Fedora might have for making the systemd journal

the

default logging option for the whole distro. (See https://lists.fedoraproject.org/pipermail/devel/2012-October/172682.html)

Of the things I listed as critical, some very key ones like time-based rotation and "journal format documented" are done. Others are still

lacking,

like a mechanism for separation of authpriv data, easy Fedora-centric disaster recovery documentation, and, simply, a lot of QA and testing.

So, the question is: do we want to be part of the testing ground for this? There are some advantages, including of course less stuff in the image, and anyone wanting more complicated configuration can easily install rsyslog, and, given limited space, automatic free space management.

Or, do we want to disable the persistent store, and rely on rsyslog?

Or, do we want to just leave it?

I think it should default to the same thing every other spin defaults to until FESCo decide to reverse their earlier decision on this issue. Differing before that happens forces people to do all the work you described, just for the benefit of a single spin that is among the easiest to customize.

-- Garrett Holmstrom