problem with cronolog and graceful restarts

[John S J Anderson <jacobs@xxxxxxxxxxxx>]

  I would like to deploy cronolog to do hourly access and error log
  rotation on a number of web servers. (We're currently doing hourly
  rotation via a cron'd script.) In my testing, however, I'm running
  into a problem.

  Basically, it appears that any long-running httpd child that is
  around during a graceful restart doesn't log properly, and doesn't
  exit. This appears to be very similar to bug 26467 in the Apache Bug
  Database ("'graceful' kills 'rotatelogs', then hangs long-lived
  httpd children"). I am testing cronolog due to this problem with
  rotatelogs.  

  Can anybody offer any suggestions for ways to solve or work around
  this problem? We need to do relatively frequent graceful restarts,
  and the rate of leakage would mean doing full restarts two or three
  times a week, which is just unacceptable. 

john.
-- 
"One must spend time in gathering knowledge to give it out richly."
  - Edward C. Steadman