|
Plans for cronolog
There are a number of things that I would like to do to improve
cronolog; many of these are listed in the wish list below.
One of the problems I have is finding much time to work on cronolog.
However I hope that creating this new web site for cronolog will give
me a structure around which to organize my cronolog activities. So
this page list some of the tasks I want to address in the short to
medium term:
- continue to tidy up the web site
- review all the emails I've had about cronolog over the years and
update the wish list and FAQ
- bring out a new release that incorporates the simpler of the
wish list items, for example
- add options --uid, --gid and --umask to
cronolog
- add an option to read from a named pipe (FIFO) rather than
standard input; this would allow cronolog to be used
with programs other than Apache.
- finish off the module version
- adapt cronolog for Apache 2.0
- make cronolog more flexibile (possibly add a configuration file)
- ... (any ideas)
Andrew Ford, 24 January 2002
cronolog wishlist
These are some of the features that are commonly requested.
cronolog as an Apache module
This is the most frequently requested feature and is something I have
been working on. Rewriting cronolog as an Apache module, would
avoid the overhead of passing all log messages through external
processes, and the overhead of having an external process per log file.
A beta version of the module is now available.
Ability to split on virtual host
This would be quite straightforward to do in an Apache module as the
module has access to information about the request, however the
standalone cronolog program does not parse the information in the
log file entries.
Running external post-processing programs
I would also like to add an option to run an external program
(such as gzip or a stats program) on each log file as
it is finished with (this was suggested by Juergen Lesny
<lesnyj@informatik.tu-muenchen.de>). However when run by
Apache there may be multiple cronolog processes running (the
external program should probably be only run once when the last
cronolog process closes the file). One idea I had to get round
this was for each process to read-lock the log file it has open
and then try to write-lock it before it closes it. If the
attempt to get a write-lock succeeds, then the process then has
exclusive access to the file and so can run the external
program; otherwise the process should just close the file and
leave it to another process to run the program. Cronolog
probably also needs a timer to signal that files that have
become out-of-date should be closed rather than just detecting
the situation when it comes to write the next log entry, which
could be several minutes, hours or even days after the end of
the period. Suggestions on this aspect are welcome.
Create new log files at rotation time
Currently cronolog waits on its standard input for more data
and only creates a new log file when data arrives after the
roll-over time. However this could be well after the roll-over time
and if entire periods go by with no data then cronolog will
not create empty log files.
Does cronolog support large file
There is no explicit support for large files. Cronolog just opens
each log file in append mode and carries on writing to it. Some
operating systems impose a limit of 2 gigabytes on the size of files.
Cronolog ought to detect when it has reached such a limit. One
possible enhancement would be to create overflow files by appending a
sequence number, i.e. create files named so:
/web/logs/2002-02.log
/web/logs/2002-02.log.1
/web/logs/2002-02.log.2
...
Of course if you are generating more than two gigabytes of log
information per rotation period, then the obvious thing to do is to
choose a shorter rotation period. You might also want to consider
separating out certain categories of file, such as images
# flag requests for image files
SetEnvIf Request_URI (gif|png|jpe?g|tiff?)$ IMAGE=1
# separate log entries into image and non-image log files
CustomLog "|/usr/sbin/cronolog /web/logs/%Y-%m.log" env=!IMAGE
CustomLog "|/usr/sbin/cronolog /web/logs/%Y-%m-images.log" env=IMAGE
Other
There are other suggestions and wish list items in the mailing lists,
which I will add here.
|
