From: Walker, Bruce J <bruce.walker@hp...> - 2005-03-20 06:35:59
1. For the most part, the SMP kernel works fine on UP machines. There
are some exceptions.
2. LVM has been used on single machines and should work with DRBD but
I'm not sure anyone has tried. However, having it present on the
failover node may be a problem. The way CFS and failover, etc. work,
you don't "mount" the drbd device on the second node. The fstab
indicates that the filesystem is node=3D1:2, which means mount it either
on node 1 or node.
3. Tools: You can run nagios and some other standard tools. Other than
webview, there isn't much OpenSSI specific (top, ps, etc. work
clusterwide, however). What would you like.
> -----Original Message-----
> From: ssic-linux-users-admin@...=20
> [mailto:ssic-linux-users-admin@...] On=20
> Behalf Of Iain Young
> Sent: Saturday, March 19, 2005 9:18 AM
> To: ssic-linux-users@...
> Subject: [SSI-users] OpenSSI, SMP vs UP, DRBD, and LVM/Device Mapper
> Hi Everyone,
> (Sending to the users list, although since this relates to linux 2.6
> and OpenSSI 1.9, it may also be appropriate to the devel list as well,
> but its more of a user issue than a devel issue)
> I'm looking at implementing an OpenSSI cluster under Linux 2.6 (Yes,
> I know that OpenSSI 1.9 isnt out yet, but I'm quite used to running
> stuff from CVS *grin* and I'm not quite ready to implement the cluster
> I've read thru the docs that are avaliable on the website, and what
> Im thinking of doing is having a couple of init nodes holding all the
> filesystems, and a number of "computational" nodes doing all the
> processing work.
> First question that comes to mind is that some of my nodes aren't
> likely to be identical hardware. Infact, some are likely to be single
> processor, and others dual processor boards.
> Is there a way to have different kernels for different nodes in the
> cluster, or do they all get the same kernel ? Obviously different
> hardware (eg netwok cards) could be handled thru modules, but SMP vs
> Uni processor is a little fundamental.
> [I guess I could just run an SMP kernel on a UP box, but I seem to
> remember health warnings about doing that]
> Second question I have relates to the file systems. As I=20
> understand it,
> the first node's filesystems are seen by all the other nodes in the
> cluster. I further understand that you can duplicate the data onto
> other nodes for failover and redundancy purposes using drbd.
> I'd like to use LVM, and the underlying device mapper to provide some
> software RAID. The logical volumes are just block devices, so I guess
> it should "just work", but can someone confirm or correct my=20
> understanding ?
> Am I correct that I would use the /dev/drbd device to mount the
> filesystem at that point on the second node ? What about on=20
> the initial
> node ?
> Third, and final question, what userspace tools are avaliable for
> viewing and managing the state of the cluster ? I know about OpenSSI
> Webview, but are there any other similar tools ?
> Thanks in Advance, and All the Best
> SF email is sponsored by - The IT Product Guide
> Read honest & candid reviews on hundreds of IT Products from=20
> real users.
> Discover which products truly live up to the hype. Start reading now.
> Ssic-linux-users mailing list