Gentoo Forums
Gentoo Forums
Gentoo Forums
Quick Search: in
cluster management
View unanswered posts
View posts from last 24 hours

Reply to topic    Gentoo Forums Forum Index Other Things Gentoo
View previous topic :: View next topic  
Author Message

Joined: 20 Sep 2014
Posts: 5

PostPosted: Sat Oct 04, 2014 10:50 pm    Post subject: cluster management Reply with quote

I have a small, and very basic, infiniband cluster running with 150 diskless nodes. The setup is very similar to the 'Gentoo disless howto' where I have the slave file system stored in it's own directory on the master nodes fs.

The problem with this configuration is that, when I need to add a new user to the system, that user also needs to be on the slave nodes file system so I copy files (/etc/passwd, /etc/group, /etc/shadow, etc...) from the master to the slave fs to make sure everything on the slave mirrors the master. The users home directory is exported from the master node so that the user has the same home dir on all nodes. And then I have to reboot the entire cluster for the changes to take effect.

Other than this issue, I do very little work aside from a few scripts to copy and run processes remotely. So I'm looking for a better way to handle adding users w/ no reboots as well as some better tools in general to move me into the next century. Any advice would be appreciated.

I also have this one other question. It's well known that if you let Gentoo sit stale w/o emerging updates for a year or more, then you're likely to run into portage problems (udev, profiles, etc...). However, no one likes downtime so I don't update for years. Unfortunately, when that time comes, its a battle. Any comments on this?
Back to top
View user's profile Send private message

Joined: 01 Jul 2004
Posts: 4004
Location: USA

PostPosted: Sun Oct 05, 2014 2:24 pm    Post subject: Reply with quote

The first question you need to look into LDAP or YP/NIS which is meant for clusters to share the same authentication information. I ended up using YP/NIS on my home network as it's very simple and supported in many UN*Xes since antiquity (and I don't have to worry too much of security implications), LDAP is the modern replacement.

As for the system reboots, yes this is a problem of Gentoo IMHO. You can't really use Gentoo to keep up to date without risk for breakage. Likely you should use a fixed revision distribution so that when critical updates come, there's little disruption, else you don't need to update. Then after two-three years you reinstall the whole OS...
Intel Core i7 2700K@ 4.1GHz/HD3000 graphics/8GB DDR3/180GB SSD
What am I supposed to be advocating?
Back to top
View user's profile Send private message
John R. Graham

Joined: 08 Mar 2005
Posts: 7836
Location: Somewhere over Atlanta, Georgia

PostPosted: Sun Oct 05, 2014 2:41 pm    Post subject: Reply with quote

Or else, if you're really determined to use Gentoo—and there are advantages—set up a much smaller test cluster to qualify your updates on. Do all your updates there and then quickly deploy the validated image to your production cluster.

The comments on LDAP are spot on: it's exactly what it's for.

- John
This space intentionally left blank.
Back to top
View user's profile Send private message

Joined: 27 Aug 2013
Posts: 391

PostPosted: Sun Oct 05, 2014 8:34 pm    Post subject: Reply with quote

NIS is definitely an option, but it might even not be necessary. A few ideas:

How is your cluster organized? I had a (much smaller) beowulf-like cluster where diskless nodes would readonly mount / from master using NFS (and then aufs to turn it rw - and go with different runlevel than master node). Whether it's good or not deppends mostly on bandwidth and security level you are interested in. I was perfectly fine with No File Security share in trusted environment, so any changes made to host would propagate to clients, unless they have already applied their own changes (and created higher priority layer for this particular file)

More secure (and most likely more less demanding in terms of bandwidth) way would be using public key authentication.

Also, there are filesystems designed to be accessed at block level by several machines at the same time, like e.g. GFS. Their performance might vary, but then again it might or might not be good enough for you.
Back to top
View user's profile Send private message
Display posts from previous:   
Reply to topic    Gentoo Forums Forum Index Other Things Gentoo All times are GMT
Page 1 of 1

Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum