Hi all,
I've been thinking about creating a local repo for our to-be CentOS-park (we're currently in the process of migrating from RHEL3 to CentOS5). As we need to control all updates being applied (think WUS, Windows Update Server) because of special software being run on servers and clients, we must check first that our molecular modeling software doesn't bite the grass should eg X, the kernel (and breaking the proprietary nvidia gfx-drivers in the process) etc be updated in a "non-approved" way.
Therefore I've planned to create a local repo on one of our servers, share /var/cache/yum, set keepcache to 1 on that server and have it reposync periodically (like once a week) with CentOS Base, Extras, rpmforge and so on.
On the clients I'd disable all standard repos, and only have the clients yum update from the aforementioned local repo server.
Would this work? Or is there a cleaner/nicer/better/simpler way to do this? Gotchas' maybe? Any feedback is appreciated.
Thx in advance.
Sorin Srbu wrote: ...
Therefore I've planned to create a local repo on one of our servers, share /var/cache/yum, set keepcache to 1 on that server and have it reposync periodically (like once a week) with CentOS Base, Extras, rpmforge and so on.
Why not make a "real" mirror using rsync, test new updates on a test test machine, and run "yum update" on the clients when the test machine is ok?
You could risk getting into trouble updating via a shared /var/cache/yum when CentOS 5.3, 5.4, etc. comes out.
Use nvidia-x11-drv from rpmforge to get updated nvidia drivers.
Mogens
Mogens Kjaer <> scribbled on Thursday, October 23, 2008 10:32 AM:
Therefore I've planned to create a local repo on one of our servers, share /var/cache/yum, set keepcache to 1 on that server and have it reposync periodically (like once a week) with CentOS Base, Extras, rpmforge and so on.
Why not make a "real" mirror using rsync, test new updates on a test test machine, and run "yum update" on the clients when the test machine is ok?
Ok, that's a point I guess. Although, I don't quite see why this is different from the setup I mentioned though, could you elaborate a bit? What would the advantage be with your suggestion?
Also, how can I let all the clients automatically update from the server, when the updates have been okayed? Yum-nightly maybe?
You could risk getting into trouble updating via a shared /var/cache/yum when CentOS 5.3, 5.4, etc. comes out.
I'd use repomanage to weed out old stuff, maybe with a cronjob, and only keep the latest versions. Also, there's always some hub-bub on this list, so I can preemtively disable the reposync, until the new version's been tested.
Use nvidia-x11-drv from rpmforge to get updated nvidia drivers.
Are they any good? I inherited this linux-farm from the previous *nixadmin, and the way it's setup assumes we use the proprietary drivers from nvidia, as that's what the molecular modeling software company says'll work. We need the fancy 3D-stuff and whatnot.
TIA.
/S
Sorin Srbu wrote: ...
Use nvidia-x11-drv from rpmforge to get updated nvidia drivers.
Are they any good? I inherited this linux-farm from the previous *nixadmin, and the way it's setup assumes we use the proprietary drivers from nvidia, as that's what the molecular modeling software company says'll work. We need the fancy 3D-stuff and whatnot.
nvidia-x11-drv is the proprietary drivers from Nvidia, just packaged so that you (normally) don't have to worry about kernel upgrades.
We use this on a CentOS machine running molecular modelling software without problems.
Mogens
Mogens Kjaer <> scribbled on Thursday, October 23, 2008 12:13 PM:
Use nvidia-x11-drv from rpmforge to get updated nvidia drivers.
Are they any good? I inherited this linux-farm from the previous *nixadmin, and the way it's setup assumes we use the proprietary drivers from nvidia, as that's what the molecular modeling software company says'll work. We need the fancy 3D-stuff and whatnot.
nvidia-x11-drv is the proprietary drivers from Nvidia, just packaged so that you (normally) don't have to worry about kernel upgrades.
We use this on a CentOS machine running molecular modelling software without problems.
That's good info, I'll look into this. Thx!
-
Mogens Kjaer -
Sorin Srbu