[Freeipa-devel] [PATCH 0434] log: add timestamp to filename of logs
Rob Crittenden
rcritten at redhat.com
Mon Mar 14 12:56:47 UTC 2016
Jan Cholasta wrote:
> On 11.3.2016 15:56, Gabe Alford wrote:
>> On Fri, Mar 11, 2016 at 7:35 AM, Petr Vobornik <pvoborni at redhat.com
>> <mailto:pvoborni at redhat.com>> wrote:
>>
>> On 03/11/2016 03:00 PM, Rob Crittenden wrote:
>>
>> Martin Kosek wrote:
>>
>> On 03/11/2016 09:55 AM, Jan Cholasta wrote:
>>
>> On 11.3.2016 09:33, Martin Kosek wrote:
>>
>> On 03/08/2016 07:07 PM, Martin Basti wrote:
>>
>>
>>
>> On 08.03.2016 16:37, Martin Basti wrote:
>>
>>
>>
>> On 08.03.2016 16:31, Martin Basti wrote:
>>
>>
>> https://fedorahosted.org/freeipa/ticket/4501
>>
>> Patch attached.
>>
>>
>> Rebased patch attached.
>>
>>
>>
>> self-NACK
>>
>> Scripts print to CLI unformatted strings, it
>> should not be so easy.
>> See /var/log/ipaupgrade-{timestamp}.log for more
>> information
>>
>>
>> second-NACK. We cannot break existing log file
>> paths. The paths are mentioned
>> in a documentation and there may be also automation
>> around that (gathering log
>> files). So there should be always symlink from the
>> well known location to the
>> newest timestampe'd log.
>>
>>
>> Sorry, but this is absurd. What's the point of
>> maintaining backward
>> compatibility with obsolete documentation? Following
>> this logic, we would not
>> be able to change anything ever. What we should actually
>> do is update the
>> documentation. Ditto for automation.
>>
>>
>> +1 for updating the automation and documentation. But some
>> backward
>> compatibility will need to be retained, at least for the
>> stable systems like
>> RHEL where *other* people may have some automation or
>> documentation around it,
>> not just us.
>>
>>
>> Or you could just also create a symlink to the old name and it
>> will
>> always just work.
>>
>> rob
>>
>>
>> Aren't the symlinks what Martin2 mentioned in second-NACK?
>>
>> These new timestamped logs should be combined with the Gabe's
>> patches: #5728 (renamed to command name) and #5724 (move to
>> /var/log/ipa directory).
>>
>> So that there will be e.g.:
>> /var/log/ipaserver-install.log ->
>> /var/log/ipa-server-install-{timestamp}.log
>>
>> /var/log/ipa/ipa-server-install.log ->
>> /var/log/ipa-server-install-{timestamp}.log
>>
>>
>> I wonder if it would be simpler/better to always write to the *.log
>> file, and then have old logs timestamped rather than write directly to a
>> timestamped log file?
>> Then just symlink the original log file in /var/log/ to the new log file
>> name/location in /var/log/ipa.
>>
>> For example:
>> /var/log/ipaserver-install.log ->
>> /var/log/ipa/ipa-server-install.log <-- We write to this
>> log (current)
>>
>> /var/log/ipa-server-install-{timestamp}.log <-- Old log with some date
>>
>> /var/log/ipa-server-install-{timestamp}.log <-- Older log with some
>> date
>>
>> /var/log/ipa-server-install-{timestamp}.log <-- Oldest log with some
>> date
>
> This is way too overengineered for something that should actually be
> really simple. I don't care if it is done this way or not, but IMHO it
> would be a waste of time. Logs are not API and should not be treated as
> such. If it needs to be done differently on RHEL, it should be handled
> downstream.
Sure logs are not API but they have been named the same way since
inception (nearly 8 years now). I don't think symlinking to the old
names is a big deal.
rob
More information about the Freeipa-devel
mailing list