[Yum] Newbie question

Tom Diehl tdiehl at rogueind.com
Sat Nov 15 14:15:16 UTC 2003


On Sat, 15 Nov 2003, Rick Thomas wrote:

> 
> On Friday, November 14, 2003, at 10:56 AM, nosp wrote:
> 
> >
> > For me, I have one NFS directory all the machines can mount, and
> > symlinked /var/cache/yum/*/packages and /var/cache/yum/*/headers 
> > to that
> > directory.  I make sure one machine does a yum check-update and yum
> > upgrade before all the others, so there is only one writer to that
> > directory (practically speaking, though there is a danger).

??

> This is a neat trick, and definitely saves some bandwidth if your 
> are at the far end of a very narrow pipe.  However, it's not for 
> the faint of heart!  A lot depends on getting the timing exactly 
> right, and that's not a trivial task.

Timing?? What is nontrivial about it??

> It's a game for trained professionals... Definitely in the class of 
> "Don't try this at home, kids!"
> 
> Disk is cheap.  Doing the default thing -- duplicating the yum 
> headers and packages per machine -- is definitely the safest plan.

What is wrong with having 1 machine that houses the repository and
all of the others accessing it via nfs/ftp/http ot whatever. Just
have the machine that houses the repo do the syncing. I agree disk
is cheap but in the above example putting the same info on multiple
disks is a waste.

I can tell you for sure that autofs/nfs/yumfailover can handle this stuff
easially and sanely. For the machines on my local network they access the
repo via nfs. For the machines that travel the failover mechenism in yum
allows them to transparntly fail over to ftp/http when they are on the road.

HTH,

..........Tom




More information about the Yum mailing list