On 16/07/12 11:26, Bjoern Boschman wrote: > Hi Nicholas, > > the basic problem can be reproduced as follows: > > configure error logging: > > [mysqld] > log-error = /var/log/mysql/error.log > > and restart mysql-server > > chown root /var/log/mysql/error.log > mysqladmin --defaults-file=/etc/mysql/debian.cnf flush-logs > mysqladmin: refresh failed; error: 'Unknown error' > > > The problem seems to be related in mysqld_safe which under some > circumstances creates the log-error file with root as file owner. > > I definately did not do this on purpose, but it might come from on > older 5.1 installation.
Well I want to know how this happened before fixing it. Otherwise I think the risks outweight the benefits. > > When you manually fix the file-owner (chown mysql > /var/log/mysql/error.log ) the problem disappers. > > -snip- > regarding mysqld startup: > > the SVN/experimental fix is basically :ugly: > from some discussion on the #mysql.de channel I very likely believe we > should not be using mysqld_safe at all, but start the mysqld process > directly. > Unfortunatelly I did not find the time to write a proposal for an > updated init-script yet :( I am somewhat more sympatheic to this view than CLint appears to be, but that does not amount to a commitment to implement it. What I want to do is as follows: 1.) work out how to reproduce or otherwise the various init/maint script bugs. If possible these should be incorporated into the integration tests I now have. 2.) Then compare the costs and benefits working with the existing set up as opposed to getting rid of mysqld_safe. (systemd and upstart also need to be considered here). Really 1.) has to be done before 2.) Also I want to the conversion to git rename the source package to "mysql" and do the package restructuring first. > > -snip_ > regarding logrotate: > > problem with logrotate basically is that we do not have any clue which > logfiles will be created as this information comes from the mysqld config. > From the distributed my.cnf no logging is done at all. Yes this is why I like this change. It looks like it enhances the robustness of our packaging. Thanks. > > I'll get back to you @IRC as soon as my employee will leave me some > time to do so ;-) Best wishes to your employer. You're welcome to work on this but a delay suits me fine . > > > Cheers > B > -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org