Re: [SSI-users] Debian Lenny OpenSSI with LVM and RAID 1

[Cumberland, L. <lon...@ni...>]

Hello John, and All,

I am looking at various solutions and implementations to test and have come across MHDDDFS (which uses the FUSE libraries) as a stackable system on mount points:

http://romanrm.ru/en/mhddfs

http://svn.uvw.ru/mhddfs/trunk/README

I was trying to load it into the main server but seem to get errors and as being afraid of crashing the cluster, I thought that I would ask the list if anyone had some ideas?

-------------------------------------------------

spartan:/mnt# apt-get install mhddfs

Reading package lists... Done
Building dependency tree
Reading state information... Done
The following extra packages will be installed:
  fuse-utils
The following NEW packages will be installed:
  fuse-utils mhddfs
0 upgraded, 2 newly installed, 0 to remove and 2 not upgraded.
Need to get 0B/38.2kB of archives.
After this operation, 213kB of additional disk space will be used.
Do you want to continue [Y/n]?
WARNING: The following packages cannot be authenticated!
  fuse-utils mhddfs
Authentication warning overridden.
Selecting previously deselected package fuse-utils.
(Reading database ... 102736 files and directories currently installed.)
Unpacking fuse-utils (from .../fuse-utils_2.7.4-1.1+lenny1_i386.deb) ...
Selecting previously deselected package mhddfs.
Unpacking mhddfs (from .../mhddfs_0.1.12-1_i386.deb) ...
Processing triggers for man-db ...
Setting up fuse-utils (2.7.4-1.1+lenny1) ...
creating fuse group...
udev active, skipping device node creation.
invoke-rc.d: WARNING: Service udev has no entry in rc.nodeinfo
invoke-rc.d: Starting only on initnode
Usage: /etc/init.d/udev {start|stop|restart|force-reload}
invoke-rc.d: initscript udev, action "reload" failed.
dpkg: error processing fuse-utils (--configure):
 subprocess post-installation script returned error exit status 1
dpkg: dependency problems prevent configuration of mhddfs:
 mhddfs depends on fuse-utils; however:
  Package fuse-utils is not configured yet.
dpkg: error processing mhddfs (--configure):
 dependency problems - leaving unconfigured
Errors were encountered while processing:
 fuse-utils
 mhddfs
E: Sub-process /usr/bin/dpkg returned an error code (1)

spartan:/mnt#
-------------------------------------------------

I would like to load up the FUSE libraries unless the currently loaded OpenSSI stuff can already handle this idea in some way since it seems that I would have to install DRBD and do not know enough about the CFS to work with it heavily yet.

Thanks and have a great day,
Lonnie Cumberland, Prof.
Physicist
 


> -----Original Message-----
> From: John Hughes [mailto:jo...@Ca...]
> Sent: Tuesday, August 10, 2010 12:31 PM
> To: Cumberland, Lonnie
> Cc: Scott Walters; Openssi users
> Subject: Re: [SSI-users] Debian Lenny OpenSSI with LVM and RAID 1
> 
> Cumberland, Lonnie wrote:
> > Now I am looking at the disk drives that are located on each node for
> mapping into a single filespace so that all space appears to be a
> single drive to the users. For this, I was reading that OpenSSI appears
> to be using DRBD (if I understand correctly) which allows it to mount
> the filespace on each node. If this is so then I am guessing that
> perhaps I could use LVM to bring all of the drives together into a
> volume.
> >
> > If that is not possible then perhaps I am utilize a stackable files
> system from FUSE like XtreemFS, gfs, or some other. Any ideas?
> >
> > I am also going to be adding RAID 1 to the system so that the main
> drive has a complete mirror and failover incase the main drive goes
> down then the secondary will pick up.
> >
> Your problem with linux raid and lvm wil be making sure that things get
> failed over from node to node in the event of a node crash.
> 
> This works:
> 
> use DRDB to mirror between the nodes
> use CFS to make the filesystem on the DRBD device available to all the
> nodes.
> 
> This also works:
> 
> Use shared disk hardware (SAS, SCSI, Fibre Channel, whatever) to make
> the disks available to all the nodes.
> Use CFS on the disks to make the filesystem to all the nodes.
> 
> This has worked in the past, I have no practical experience:
> 
> Use a cluster aware filesystem (Lustre, even NFS) and mount the
> filesystem on all the nodes.
>