[redhat-lspp] tmpwatch

[Linda Knippers]

I suggest starting with the simple man page update that educates
the admin.  It should also be documented as something to be
aware of wherever the polyinstantiation setup is documented.

-- ljk

Janak Desai wrote:
> 
> tmpwatch is an administrative tool that allows removal of files that
> haven't been accessed for a period of time. tmpwatch recursively
> traverses a directory to remove temporary files. It can be used for
> any directory, however it is most often used for directories such as
> /tmp and /var/tmp.
> 
> If polyinstantiation is setup for the admin account as well as for
> regular users, tmpwatch will not traverse instances of other users.
> I can think of two possible approaches to address this, and I would
> appreciate if you could share your thoughts on them or suggest any
> other way to handle this.
> 1) Update the man page for tmpwatch with a blurb about polyinstantiation
> and instruct the admin to unmount the polyinstantiated directory before
> executing tmpwatch on it. Basically educate the admin and don't try and
> automate any unmounting through tmpwatch.
> 
> 2) Add an option to tmpwatch that will automatically unmount admin's
> instance before recursively traversing the target directory
> to remove temporary files. With this approach tmpwatch will have to
> become PAM-aware and would need appropriate PAM configuration files.
> While this approach appears to be more admin friendly, it can get
> useless and confusing if instance directories are not setup to be
> subdirectories of the polyinstantiated directory. For example,
> this option would be useful if polyinstantiation of /tmp is setup as
>                         tmp
>                  /   \
>                 /     \
>      tmp.inst-root  tmp.inst-user-joe
> 
> Then, when root executes 'tmpwatch /tmp' with an option to traverse
> instances, both tmp.inst-root and tmp.inst-user-joe would be cleaned up.
> However, if polyinstantiation is setup in such a way that instance
> directories are not subdirectories of the polyinstantiated
> directory like
> 
>               /
>            /    \
>           /      \
>          tmp    instances                  /   \
>                 /     \
>      tmp.inst-root  tmp.inst-user-joe
> 
> Then 'tmpwatch /tmp' with an option to traverse instances will only
> delete files in /tmp. 
> Basically, even if we provide a new option to tmpwatch, the admin
> will have to have a good understanding of polyinstantiation and
> its setup. I would prefer to educate the admin rather than try
> to put intelligence in the command which may or maynot help. Of
> course, it also helps that there is less work in option 1) :-).
> Please let me know what you think.
> 
> -Janak
> 
> 
> -- 
> redhat-lspp mailing list
> redhat-lspp at redhat.com
> https://www.redhat.com/mailman/listinfo/redhat-lspp
>