On 7/28/2015 4:28 AM, Dmitry Tantsur wrote:
On 07/27/2015 02:11 PM, Sean Dague wrote:
Honestly, I think deprecating and removing 'verbose' is probably the
best option. INFO is probably the right default behavior, and it's not
really "verbose" in any real openstack usage. It is unlikely that anyone
would want that to be in the off state, and if so, they can do that via
python logging config.
Thanks!
As I saw no objections, I went ahead and proposed
https://review.openstack.org/#/c/206437/
On 07/27/2015 06:32 AM, Dmitry Tantsur wrote:
Hi all!
I didn't find the discussion on the ML so I feel like starting one.
What was the reason of setting verbose to false by default? The patch
[1] does not provide any reasoning for it.
We all know that software does fail from time to time. While the default
level of WARN might give some signal to an operator that *something* is
wrong, it usually does not give much clues on *what* and *why*. Our logs
guidelines define INFO as units of work, and the default means that
operators/people debugging their logs won't even be able to track
transitions in their system that lead to an error/warning.
Of all people I know 100% are using DEBUG level by default, the only
post I've found here on this topic [2] seems to state the same. I
realize that DEBUG might give too much information to process (though I
always request people to enable debugging log before sending my any bug
reports). But is there really a compelling reason to disable INFO?
Examples of INFO logs from ironic tempest run:
ironic cond:
http://logs.openstack.org/62/202562/7/check/gate-tempest-dsvm-ironic-pxe_ssh/090871b/logs/screen-ir-cond.txt.gz?level=INFO
nova cpu:
http://logs.openstack.org/62/202562/7/check/gate-tempest-dsvm-ironic-pxe_ssh/090871b/logs/screen-n-cpu.txt.gz?level=INFO
and the biggest one neutron agt:
http://logs.openstack.org/62/202562/7/check/gate-tempest-dsvm-ironic-pxe_ssh/090871b/logs/screen-q-agt.txt.gz?level=INFO
As you can see, these logs are so small, you can just read through them
without any tooling! Of course it's not a real world example, but I'm
dealing with hundrer-of-megabytes debug-level text logs from nova +
ironic nearly every day. It's still manangeable for me, grep handles it
pretty well (to say nothing about journalctl).
WDYT about changing this default on oslo.log level?
Thanks,
Dmitry
[1] https://review.openstack.org/#/c/18110/
[2]
http://lists.openstack.org/pipermail/openstack-operators/2014-March/004156.html
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe:
[email protected]?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: [email protected]?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
This should be cross-posted to the openstack-operators mailing list
since it will impact operators and please be sure that something gets
into the Liberty release notes about this change. Most projects are
pretty good about not logging a ton of stuff at INFO level but there are
others that think everything is super important to everyone and are too
verbose.
--
Thanks,
Matt Riedemann
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: [email protected]?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
