[torqueusers] torque daemons can not be stopeed during uninstall

Garrick Staples garrick at usc.edu
Tue Dec 9 14:39:59 MST 2008


On Tue, Dec 09, 2008 at 04:32:39PM -0500, Steve Young alleged:
> 	No I don't use rpm versions of torque server because too many times  
> you get all sorts of dependancy problems. I'd rather compile it  
> against the libs I currently have installed on the server.

Again, we are talking about rpms that you are building.  You can not build rpms
against libs that aren't on your own machine.


>  I do expect that for rpm packages (that I do use), when I tell it to  
> remove the package that it actually does stop the daemon and remove  
> the package. That seems to be the way most rpm's work. I would be a  
> little confused to remove the rpm and find that afterwards there is  
> still the daemon running? I could see using an rpm package for the  
> client nodes (pbs_mom).  There I would still expect that a uninstall  
> would stop the daemon and then remove the package.

How about a compromise?  How about printing a warning message?

 
> 	For me, It's much easier to compile the app and when I need to make  
> changes/upgrades I can compile another version and a quick stop of the  
> daemon and start of the new daemon will get me running. If there are  
> problems I can quickly revert back to the previous version. Suppose I  
> remove the rpm package and try to install a new version only to find  
> out that because of dependancies that I can't install the new version.  
> Now I'm stuck and I hope I can re-install the old version of the  
> package and get everything configured back the way I had it. This is  
> why I don't like to use rpm's for certain server applications.

Again, you *are* compiling the app.  CRI doesn't distribute binary rpms.

And that stuff about deps is silly.  Rpms are upgraded with 'rpm -U' which does
the uninstall of the old package *after* installing the new package.  You can't
remove a package because you don't have the deps for the new package.

Of course, you are free to use whatever you want.  I'm not actually saying that
you should be using rpms.  I would just hate for you to make that decision with
bad information!
 

> 	I do understand your point Garrick, and I agree with you about  
> torque. It can be a tricky beast sometimes. I'm just pointing out that  
> as an Admin I would actually expect that an rpm package would stop the  
> daemon and remove it.  Why would anyone not want it removed if that's  
> what they asked for?

Well, there are 2 possible scenerios.

*shrug*

-- 
Garrick Staples, GNU/Linux HPCC SysAdmin
University of Southern California

See the Dishonor Roll at http://www.californiansagainsthate.com/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://www.supercluster.org/pipermail/torqueusers/attachments/20081209/2e23094d/attachment.bin


More information about the torqueusers mailing list