BIND 10 #1518: 0 byte rotated log files when changing settings

[BIND 10 Development]

#1518: 0 byte rotated log files when changing settings
-------------------------------------+-------------------------------------
                   Reporter:  jreed  |                 Owner:  UnAssigned
                       Type:         |                Status:  reviewing
  defect                             |             Milestone:
                   Priority:         |  Sprint-20120124
  critical                           |            Resolution:
                  Component:         |             Sensitive:  0
  logging                            |           Sub-Project:  Core
                   Keywords:         |  Estimated Difficulty:  7 (but see
            Defect Severity:  High   |  comment)
Feature Depending on Ticket:  none   |           Total Hours:  0
        Add Hours to Ticket:  0      |
                  Internal?:  0      |
-------------------------------------+-------------------------------------

Comment (by vorner):

 Hello

 > 2. Even if (1) were not a problem, we can expect processes to end up
 logging to different log files when the log file size reaches the maximum
 limit.

 This sounds possible. Also, this might have happened at the time it did
 the reconfiguration, if it bufferes a message or two, the chance that the
 limit is reached at the time the loggers dump the buffer is higher than
 usual.

 > '''Suggestions'''[[BR]]
 > 2. I think the second problem could be mitigated by using a class
 derived from the log4cplus !RollingFileAppender that overrides some of its
 methods.  When a rollover happens, instead of closing the current file,
 rolling the file names and opening a file, it should close the file, open
 what it thinks should be the current file and try to write again.  Only if
 this file exists and is at maximum size should the rollover be performed.
 We might also want to consider the idea of a lock file to prevent two
 processes trying to rollover a particular log file at the same time. (In
 fact, use of a lock file might well solve the first problem as well.)

 It might be so, I guess it would be also the fastest way to test the
 theory, as it shouldn't be so much code. And I agree with Jelte it could
 solve (1) as well. Maybe the library wasn't meant for this kind of use.

-- 
Ticket URL: <http://bind10.isc.org/ticket/1518#comment:9>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development