On Wed, Feb 21, 2007 at 06:32:02AM -0600, Chris wrote: > On Wed, 21 Feb 2007, Jens Elkner wrote: > > > Because I wanna avoid needless downtimes: > > When patches are installed and drivers are patched, usually a reboot is > > suggested/required via patch metadata. Since the internals of a patch like > > what has been done/what influence has it one the system or what impact > > would it have on the system, if one doesn't install it are usually not > > docuented or at least hard to find or hardly understandable for a > > none-expert > > So the usualy option for a solaris user is, > > a) install the patch and reboot or die OR > > b) don't install the software, you do not need... > > Wouldn't this only come into play if the driver is actually in use? If > SUNWCreq installs FC drivers and never uses them, patching those drivers > should not affect a running system.
That should be the case, but the patch say: "reboot|reconfigure the machine" and since usually one doesn't know, what is effected by the patch and also haven't found a way to deduce the kernel module names wrt. filenames, it is probably better to do, what patch metadata says ... IMHO the patchadd should be able to find out that dynamically - can't be that hard ... > However, the odds are that when you > apply a patch cluster to a host reboots will be required anyways. Yepp. > I could see an argument against installing a software package if there was > a security issue, like a suid executable you don't want, or a service you > don't need. But if a driver sits on the host unused, and takes up maybe > 200k of disk space is it really worth the effort of removing the driver > and documenting your changes against the normal baseline install? Yes - it is worth, because I wanna avoid booting servers every 2 days ... Regards, jel. -- Otto-von-Guericke University http://www.cs.uni-magdeburg.de/ Department of Computer Science Geb. 29 R 027, Universitaetsplatz 2 39106 Magdeburg, Germany Tel: +49 391 67 12768
