Our OSU intern Lei Chai has produced a screen cast for the openSolaris work she completed for NFS/RDMA; Check it out here
Technorati Tags: nfs, opensolaris, NFS/RDMA
|
|
|
|
The current draft for NFSv4.1 (21) is out, and one of the places you can fetch it from is here
I've updated "The Picture" .. for sparse files based layouts..
Technorati Tags: nfs, nfsv4, opensolaris, pNFS, solaris
|
|
|
|
Here is an updated files layout picture that is consistent with draft-14 of NFSv4 Minor version 1.
[posted with ecto]
Technorati Tags: nfs, nfsv4, opensolaris, pNFS, solaris
|
|
|
|
While looking at some of the changes for draft-13 of NFSv4 Minor version 1, I put together a couple of diagrams that represent the pNFS layouts and devices.. All you pNFS fans.. enjoy.. Here is the Layout to Device relationship, as we would use it in LAYOUTGET :
and Here Devices returned in GETDEVICELIST:
Technorati Tags: nfs, nfsv4, opensolaris, pNFS, solaris
|
|
|
|
It was a nice sunny time (mostly) in Ann Arbor for the citi
pNFS focused 14th bake-a-thon.
|
|
|
|
I'm bit of a machead and so when i take notes i use onmi outline, when i plan a project i use omni plan, and for diagrams i use omni (nope not draw) graffle ;) -- Now why haven't they done an omni blog app ?
|
|
|
|
For those adventurous types the following source code patch will allow you to
alter the stripe width.
--- /nfs41-proto/usr/src/uts/common/fs/nfs/nfs4_state.c
+++ /issue_102/usr/src/uts/common/fs/nfs/nfs4_state.c
@@ -23,7 +23,7 @@
* Use is subject to license terms.
*/
-#pragma ident "@(#)nfs4_state.c 1.52 07/07/03 SMI"
+#pragma ident "%Z%%M% %I% %E% SMI"
#include <sys/systm.h>
#include <sys/kmem.h>
@@ -4267,6 +4267,8 @@
}
}
+int mds_default_stripe = 32;
+
mds_layout_t *
mds_gen_default_layout()
{
@@ -4283,7 +4285,7 @@
return (NULL);
args.lo_arg.loid = 1;
- args.lo_arg.lo_stripe_unit = 32 * 1024;
+ args.lo_arg.lo_stripe_unit = mds_default_stripe * 1024;
rw_enter(&mds_layout_lock, RW_WRITER);
setting the value in /etc/system to alter:
set nfssrv:mds_default_stripe = 64
--
happy pnfs-ing..
|
|
|
|
One thing that may not be very apparent with the prototype code is that
the MDS automatically generates a default layout based on the data-servers
that have 'reported in' with it. The stripe count is the known data-servers and
a default value of 32k for stripe width. This information is lost each time the
MDS is rebooted.
Technorati Tags: nfsv4, OpenSolaris, pNFS, Solaris
|
|
|
|
So dear reader, it is indeed tomorrow.
I'm thrilled to announce, I (that is me Robert) will give away a brand new
never used ipod shuffle bought with my own $'s right out of the
Austin Domain Apple store.
What's the catch ? -- well, you have to be able to panic the openSolaris pNFS
MDS server, caused by an issue that we have not already disclosed .. and you
have to provide a suggested fix.. So, that's easy eh ? ... go forth reader and
start a crashing.. -- Oh, send me email once you have done it....
robert <dot> gordon <at> Sun <dot> com
Oh, one more thing.. my team mates can not participate..
Hint: start here
Technorati Tags: ipod, nfs, nfsv4, OpenSolaris, pNFS, Solaris, ZFS
|
|
|
|
pNFS wants YOU, yes YOU! -- It longs for your attention and loving keystrokes..
Seriously.. go to the download page download, install it.. use it...
You want an incentive ? ... really ? Hmmm okay.. come back here tomorrow and i'll tell you what it is..
Robert.
Technorati Tags: ipod, nfs, nfsv4, OpenSolaris, pNFS, Solaris
|
|
|
|
A Solaris NFS/ZFS Appliance
Storage at home is always a problem. I mean, you’ll never know if that nut, bolt and whatchamacallit will come in handy one day.... The same goes for the digital age, that file, that picture... you may just need it one day, or perhaps you’d just like a place to back things up to..
I kept buying single ‘portable’ hard drives as i overflowed my various configurations.. (i have 5 Macs, and an iWill zmaxdp) It was time to have a more viable solution.. So I set out to build something that would hold about 8 disks, have an decent ethernet connection, something that could be headless..
After various configurations of motherboards and cases, Jeff Smith hit me with the clue bat, “hey, just use an Ultra 40!”. So here is the recipe:-
I Ordered an Ultra 40 with the additional hard-drive backplane kit (X4213A), the other CPU (X4191A-Z), and an extra 2 Gig of memory (X5287A-Z). Now, where will I find the all important SPUD, nope it’s not a potato, it’s the bracket that will allow a disk to be mounted in the Ultra 40 disk bays. It seems that part number 540-3024 works ! and you can find them here. How about some disks ? We like seagate, and so sprinkle some of these. Now, I said headless, well we need one of these. Next all you gotta do is assemble, configure ZFS/NFS and start using it..
Technorati Tags: nfs, nfsv4, OpenSolaris, ZFS, Solaris
|
|
|
|
I was just messing about with snv_23 and FC4; and i noticed that
if i forgot to add the no_root_squash option the Slolaris Client
would pause for a short while at the mount(1M) command and
then error out with 'resource temporarily unavailable' -- one more
thing to watch out for..
[posted with ecto]
Technorati Tags: nfs, OpenSolaris, Solaris
|
|
|
|
Since there had been some confusion over a Solaris Client failing to mount from a Linux based server (there may be a couple of issues you may bump into) -- I thought i'd share some hints and things to look at..
First, of course, your mount may fail because the item you wish to mount has not been exported from the server; using showmount -e <server> will return a list of exports for that server, if the item is not listed then you should contact the server sysadmin for help (or export it yourself :) using the exportfs command with possible updates to the /etc/exports file)
You may see a permission denied message, and if you are using snv_22 to snv_24 bits that could be due to the client using a non-reserved port. A couple of things can be done to over come that problem;
/foop *(rw,nohide,insecure,no_root_squash,sync)
see this to do that!
You may also see a problem with no such file or directory (ENOENT).
The linux NFS server implements the NFSv4 pseudo filesystem (pFS) as a separate namespace to that of the server, and hence NFSv2/v3 (since in v2/v3 we expose the servers native namespace). The root of the pFS namespace is designated in the exports file via the fsid=0 option.
So for example lets say you have some data that you'd like to share in the following directories:-
/export/proj; /export/archive; /export/www; /export/temp.
An example /etc/exports file might look like this :-
/export/proj *.dev.dot.com(rw,insecure,sync)
/export/archive *(ro,insecure,sync)
/export/www @www(rw,insecure,sync)
/export/temp *(rw,insecure,sync)
Mounting using NFSv3 our mount command would look like this :-
# mount -o vers=3 linux_server:/export/proj /proj
This will work fine for v2/v3 mounts but fail for NFSv4. The pFS is customizable by the sysadmin allowing the flexibility to present a different namespace; for our example we may choose to specify that the root for the pFS be /export; Adding the following line to /etc/exports will allow us to do that :-
/export *(rw,fsid=0,insecure,no_root_squash,sync)
In doing this we are now presenting the pFS namespace to NFSv4 Clients as :-
/proj
/archive
/www
/temp
and as such we will need to alter the mount command on the NFSv4 Client:
# mount -o vers=4 linux_server:/proj /proj
See Also:
Using NFSv4
Linux NFS Client
_____________________________
Technorati Tags: OpenSolaris Solaris nfs
[posted with ecto]
|
|
|
|
In Solaris nevada build 26; I've switched the default port for the NFS client to
use reserved (aka privileged) ports, effectively disabling 6185950 until
6319735 is putback .
|
|
|
|
It was pointed out that a Linux NFS server may export a filesystem using the insecure option.
For FreeBSD it looks like there is a sysctl(8) "vfs.nfsrv.nfs_privport" that dictates the behavior or adding the -n option to mountd(8);
|
|
|
|
This theme brought to you by the Sun Green Team.
Go Green!
I find these diagrams inscrutable but the colors a...
Excellent! -- At least there is _some_ value ;-)