On 7/30/19 3:36 PM, Richard Guy Briggs wrote:
On 2019-07-30 15:06, Lenny Bruzenak wrote:
> On 7/29/19 4:32 PM, Richard Guy Briggs wrote:
>> It is being ignored because that kernel command line extension to the
>> original feature was never backported to RHEL7.
> That would definitely do it.
>
>> In hindsight, that would have been pretty useful without causing much
>> risk. Normally feature backport is driven by customer demand. There
>> was a bit of pushback when it was first introduced upstream, but this is
>> exactly the scenario I envisioned where it would be most useful. It is
>> possible to compile your own kernel and change the default value, but
>> that's obviously a hurdle for most.
> It would definitely have been useful, some might say even necessary,
> given the audit event startup noise occurring with systemd.
Yes, this was yet another difficulty that arose with the change to
systemd from rhel6 to rhel7. The intent was to solve it first in fedora
when it switched to systemd to address this since the number of startup
messages jumped from manageable within the default backlog size to
almost double. There are also other improvements upstream that remove
some of the doubt about exactly how many log messages were lost.
> Wow. Thanks Richard, I appreciate the answer on this.
It is all there in fedora and RHEL8, so that is one possible route. It
is a bit late in the RHEL7 life cycle to commit to it, but not
impossible...
Thanks Richard and I do appreciate the insight.
For some it might be possible to switch OS baselines effortlessly,
others (including my group) it isn't.
I'm surprised other RHEL 7 consumers are not squawking; I wonder if they
do not appreciate what they are not seeing? Or perhaps they are not
starting as many services early in the boot sequence and therefore
getting that one?
For people who care, I'd say that examining the stats ("auditctl -s")
after startup would be worthwhile to see if they are losing events. Even
if on fedora or RHEL8, I guess if the default is still 64 they could
also be dropping relevant events they might want.
I know this isn't a new thing, and I should have been more diligent
myself, just saying.
LCB
--
Lenny Bruzenak
MagitekLTD