[OpenAFS] Proposed changes for server log rotation

Michael Meffie mmeffie@sinenomine.net
Fri, 03 Dec 2010 10:54:04 -0500

Jeffrey Altman wrote:
> On 12/3/2010 9:38 AM, Michael Meffie wrote:
>> Simon Wilkinson wrote:
>>> On 3 Dec 2010, at 03:33, Jeffrey Altman <jaltman@secure-endpoints.com>
>>> wrote:
>>>> Since 1.6 is disabling weekly restarts by default, this problem
>>>> should be addressed before its release.
>>> Maybe 1.6 should make logging to syslog the default?
>> Perhaps. If so, I suggest there is a need to provide an option to have
>> the servers log to file if desired, and that option could specify
>> whether the log file should be renamed on startup or not. This would
>> neatly address the original question of this thread, which is how to
>> avoid the filelog data being lost by back to back fileserver
>> restarts/crashes, and how to better support sites that want to use file
>> based logging.
> Unfortunately, this option fails to work with "bos getlog" and does
> nothing for the sites that wish to use use file logs.  Periodic restarts
> is really not a desirable recommendation.  While I believe that
> defaulting to syslog probably is the right way to go, I think we need to
> consider the impact on documentation and end user best practices.
> At the very least I believe that we need to implement an internal log
> rollover mechanism based on either max size or max time with a max
> number of old logs to be maintained.  Implementing this should not be
> overly complicated if we agree that this internal file logging is not
> meant to replace full featured log rotation tool chains.

Yes, that does make sense. This would also fix the problem of losing
filelog info on back to back restarts. Instead of blindly renaming the
log file on a restart, the server could do such only when a certain
size threshold is exceeded. This would be compatible for existing
sites while not precluding the modern tools.

Thank you Jeff,