You can use Drobos without any special software. The blue capacity lights will
fill up as the disks fill. When it gets too full, Drobo will ask for another
drive with its drive lights. Just feed it drives when it asks, and that is all.
So what is the dashboard good for?
So that is where drobo-utils, the linux dashboard, comes in.
Drobo-utils is a set of linux tools to query and manage Data Robotics
Drobo storage systems. Fire up drobom view, and it should look
familiar to those who have seen the dashboard on other operating systems.
Drobom view is built on a little python programmer interface which can be installed
on the system and used by other applications as well. For experienced Linux hands,
the rest of the command line interface is provided by other sub-commands of
drobom, and offer the same functionality as the view graphical interface.
If you can get your drobo device to be associated with a /dev/sdX
file, then you will be able to partition, build file systems and
read & write data to it using ordinary system tools (see Setup.)
Drobo-utils accesses drobos via the same files. The software
scans those files, and asking each device if it is a Drobo.
Unfortunately, the way Drobos respond varies, so not all of them respond
in a way that the software understands. Even for the same
device, different physical interconnects may work with different functionality.
There are two levels of access to Drobos: data (being able to read and
write data, and full meaning that the drobo responds to the full
command and control language.
With that in mind, the compatibility matrix of each device vs. the
physical channel is below:
Model |
Interface
( Performance MB/s ) |
Maximum
LUN Size |
interf |
USB |
FW |
TCP/IP |
iSCSI |
eSATA |
ext3 |
Drobo
(Gen1) |
full
(15) |
n/a
(0) |
n/a
(0) |
n/a
(0) |
n/a
(0) |
2 |
Drobo
(Gen2) |
full
(15) |
full?
(?) |
n/a
(0) |
n/a
(0) |
n/a
(0) |
2 |
Drobo
Share |
n/a
(0) |
n/a
(0) |
data*1
(15) |
n/a
(0) |
n/a
(0) |
2 |
Drobo
Pro |
full
(15) |
?
(?) |
n/a
(0) |
full*2
(80) |
n/a
(0) |
8-? |
Drobo
Elite |
?
(?) |
n/a
(0) |
n/a
(0) |
?
(?) |
n/a
(0) |
8? |
- Drobo
- S
|
full
(15) |
?
(?) |
n/a
(0) |
n/a
(?) |
data
(0) |
8 |
full - Find/Full: drobo-utils will find the device on it's own (auto-discover)
data - works: device functions for data i/o, but Drobo-utils cannot access it for configuration.
n/a - not applicable (device does not have this interface.)
*1 - Droboshare is not usable with drobo-utils on a linux server. If you
install Droboshare Augmented Root File System (DARFS) then one can
run a drobo-utils in line mode on the droboshare itself.
*2 - Will not detect, out of the box, an iSCSI drobo. One must configure
the iSCSI subsystem to obtain a /dev/sdX file. see iSCSI Setup for
details on that initial configuration. After that point, Drobo-utils
will function properly over iSCSI.
People sued a national fast food chain because their coffee was hot, but did not
have a warning on the cup stating that. For most people, the risk of scalding
should be fairly obvious.
Drobo has made it so much easier to obtain much more secure storage, that some with
little or no professional storage management experience are getting Drobos.
Some have the expectation that Drobo´s, because they allow for disk failures,
replace the need for backups. Sad stories have been told about people putting
data only on the drobo, and no-where else, and then something happens and they
lose the data.
You need to look at your data and determine the backup/data security strategy.
If you have never done this, or do not know what it means, please consult the
Deployment section in this page for examples.
On Ubuntu 9.10 or later (or Debian unstable or other debian derived
distributions), drobo-utils is included in the repositories, and installation
from a shell prompt is simply:
% sudo apt-get install drobo-utils
to run at least the command line utility. Users on servers often want only
command line functionality. On the other hand, to enable the graphical user
interface, one more package must be installed:
% sudo apt-get install python-qt4
That is the easiest installation method, this method ensures that any packages
required are automatically installed on the system as part of the above
installation. On other distributions, or if the version in the repositories is
too old, more complicated methods might be needed. For all other installation
methods, one must ensure the packages that drobo-utils requires are installed.
These packages are called Dependencies.
Before one can install drobo-utils itself, the other packages needed are something
like those below (these examples are ubuntu packages, names may vary on other
distributions):
python -- interpreter for python language
parted -- partitioner, usually included with the distro.
If using Redhat Enterprise Linux (RHEL, aka. CentOS, Scientific Linux etc...),
which have python 2.4 , then the following are necessary:
python-ctypes -- module for C-interface
On RPM-based distros (such as Redhat & SuSe), it would more likely be 'yum' instead of
'aptitude' and some of the package names will change. A typical difference is that
packages for developers have the -devel suffix on Redhat derived distributions, instead
of the -dev favoured by debian derived ones.
Here is an example from fedora 7 (courtesy of help4death on the google group):
% yum install python
% yum install PyQt4
% yum install python-devel
NOTE: if X or QT is missing, it will only disable the GUI. Line mode will
work without issues. The package should work fine on headless servers using
only the command line.
Once dependencies are satisfied, one can install the latest stable package manually.
Point a browser at: http://sourceforge.net/project/showfiles.php?group_id=222830
where the most current packages are available. after downloading a .deb, it is simply a matter of:
# dpkg -i drobo-utils-<version>.deb
done!
Redhat/Fedora users. alien may be used to convert the package. I don't know if it works. Someone to take on RPM packaging would be very welcome! This is a pure python package, so the chances are good that it does work without issue.
There is a Fedora Package
Assuming the dependencies are installed/satisfied, the package will actually run fine
without being installed in any systemish places. Source code can be directly downloaded
run it explicitly from the directory.
Point a browser at: http://sourceforge.net/project/showfiles.php?group_id=222830
download the .tgz preferred, then unpack it:
# tar -xzvf drobo-utils-<version>.tgz
# cd drobo-utils-<version>
# ./drobom status
for all of the examples in the manual one just needs to prepend './' before drobom.
When a new model comes out, or the stable version is missing a feature, one may elect to
follow along with the latest development version. installation of git, is necessary,
then use it can be used to get a copy of the source tree:
# apt-get install git
# git clone git://drobo-utils.git.sourceforge.net/gitroot/drobo-utils/drobo-utils
# cd droo-utils
# ./drobom status
# git pull
This gives a read-only copy of the source code that can be updated with the latest
changes with 'git pull'. One can also select any stable version of drobo-utils by use of
'git branch -r', and 'git checkout'. For details, consult git documentation.
So, one way or another, drobo-utils is installed. The next step is to try it out.
The first item to verify after installation is to invoke the
command line interface (CLI.) and see if something sensible happens...
on my system with a drobo[#sdz] the following happens:
% sudo drobom status
/dev/sdz /drobo01 100% full ( ['Red alert', 'Bad disk', 'No redundancy'], 0 )
%
very scary, but my drobo is in bad shape right now... normal result is: []
as a status, which means there is nothing wrong. To get all kinds of
information about the drobo, try 'drobom info.' Invocation
without arguments at all which will cause it to print out a list of the
commands available through the command line interface.
Once the command line functionality is verified, and assuming
python-qt4 is installed:
# drobom view
as root starts a GUI for each drobo attached to a computer.
There are various tabs to allow one to obtain information from the Drobo,
and also change its configuration. For example, one can use the Format
tab of the GUI to partition the device and create a single file system
for a given LUN.
There are two choices to make when setting up a Drobo: file system type, and LUN size.
For a discussion of these choices, please consult: LUNSIZE fits all
and Filesystem Choice.
This section provides an illustrative example of the most common configuration.
An ext3 file system is built on a Drobo with whatever LUNSIZE is already in place.
The GUI and line modes produce exactly the same result, and simply
execute standard linux partitioning using parted, and the appropriate
file system builder for the type in question. Sample CLI run:
PleaseEraseMyData# drobom -d /dev/sdz format ext3
peter@pepino:~/drobo/drobo-utils$ sudo ./drobom format ext3
/dev/sdz - Drobo disk pack 00% full - ([], 0)
preparing a format script for a ext3 file system as you requested
OK, I built the script but nothing is erased yet...
You can have a look at it with: cat /tmp/fmtscript
If you are really sure, go ahead and do: sh /tmp/fmtscript
WARNING: Ready to destroy all your data. Continue? (y/n) n
Phew... You stopped just in time!
peter@pepino:~/drobo/drobo-utils$
# cat /tmp/fmtscript
#!/bin/sh
parted /dev/sdz mklabel gpt
parted /dev/sdz mkpart pri ext3 0 100%
parted /dev/sdz print; sleep 5
mke2fs -j -i 262144 -L Drobo01 -m 0 -O sparse_super,^resize_inode /dev/sdz1
The above sets up the drobo as one big partition, with a label that says
it ought to contain an ext3[#mke3fs] file system. For an NTFS file system,
write ´ntfs´ in place of ext3. The next step is to add the file
system into the partition. while parted's are instantaneous, the mke2fs
takes a while, just have a little patience, it´ll be fine.
The ´coffee is hot´ disclaimer is necessary at this point:
WARNING: THE FOLLOWING LINES WILL ERASE ALL DATA ON YOUR DROBO!
WARNING: NO, IT WILL NOT ASK ANY QUESTIONS!
WARNING: ASK YOURSELF, before you start: ARE YOU SURE?
WARNING: AFTER THE SECOND LINE, YOU ARE TOAST.
WARNING: BEST TO BACKUP YOUR DATA BEFOREHAND...
sh -x /tmp/fmtscript
(For an ntfs file system, use mkntfs -f -L Drobo01 /dev/sdz1
... For ext3, be prepared to wait[#mkext3time]_ )
Sample run:
root@alu:~# parted -i /dev/sdz
GNU Parted 1.7.1
Using /dev/sdz
Welcome to GNU Parted! Type 'help' to view a list of commands.
(parted) mklabel gpt
(parted) mkpart pri ext2 0 100%
(parted) quit
root@alu:~# fdisk /dev/sdz
GNU Fdisk 1.0
Copyright (C) 1998 - 2006 Free Software Foundation, Inc.
This program is free software, covered by the GNU General Public License.
This program is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE. See the GNU General Public License for more details.
Using /dev/sdz
Command (m for help): p
Disk /dev/sdz: 2199 GB, 2199020382720 bytes
255 heads, 63 sectors/track, 267349 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Device Boot Start End Blocks Id System
/dev/sdz1 1 267350 2147488843 83 Linux
Command (m for help): q
root@alu:~# mke2fs -j -i 262144 -L Drobo01 -m 0 -O sparse_super,^resize_inode /dev/sdz1
mke2fs 1.40.8 (13-Mar-2008)
Filesystem label=Drobo01
OS type: Linux
Block size=4096 (log=2)
Fragment size=4096 (log=2)
8388608 inodes, 536870886 blocks
0 blocks (0.00%) reserved for the super user
First data block=0
16384 block groups
32768 blocks per group, 32768 fragments per group
512 inodes per group
Superblock backups stored on blocks:
32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,
4096000, 7962624, 11239424, 20480000, 23887872, 71663616, 78675968,
102400000, 214990848, 512000000
Writing inode tables: done
Creating journal (32768 blocks): done
Writing superblocks and filesystem accounting information: done
This filesystem will be automatically checked every 26 mounts or
180 days, whichever comes first. Use tune2fs -c or -i to override.
root@alu:~#
root@alu:~# mount /dev/sdz1 /mnt
By default, Drobo creates a separate 'disk' visible to the computer for every 2
Terabytes (TiB) of parity-protected capacity on the unit. The natural usage
that a drobo invites in users is to have a single, large device covering all the
data on device. For example, on Mac OS/X, users often create 16 TB LUNS on HFS.
This allows all the storage to fit on one large pool. The downside of larger
LUNS has to do with startup time, and the time to perform a file system check.
Under Linux unfortunately, with a first generation Drobo, one should limit the
volume size to 2 TiB[#gen12TiB]_. It is hoped, but not confirmed, that later
products support LUNS larger than 2 TiB on Linux. Drobom view therefore limits
lunsize to 2 TiB for the moment. The command line interface can be used to
create larger LUNS, they just might not work.
- ON LUNSIZES >= 2TB:
- -- On older distributions, there are a couple of gotchas related to
- linux tools which aren't 2TB ready... to exceed 2 TB, you need to:
- -- use GPT partitions, which aren´t supported by older fdisk
- versions. Tools based on libparted work fine, mostly.
- -- gparted fails, and seems to have a 1 TB limit on devices.
- (bug #524948 reported to bugzilla.gnome.org) It's just the GUI,
as libparted is fine, and other tools based on it
still work.
- -- on linux kernel < 2.6.24, the USB layer won't let one address
- LUNs/offsets > 2 TB. For example, Ubuntu hardy (8.04) released in Spring
2008 has a 2.6.24, and so is OK. I've never been able to test this problem.
- -- On linux kernel < 2.6.31 there is are reported firewire problem that will
- prevent devices > 2 TiB from working.
- -- ext3 with 4K blocks is supposed to allow file system capacity of 8 TiB.
- 4K blocks seem to be assigned by default. So I think a good max.
It would be fun to set the LunSIZE to 8 TiB and test it out...
-- Windows XP does not support LUNS > 2 TiB
Drobos work well under linux with ext3. One can, of course, set up an NTFS or
HFS+ or FAT32 if necessary, but it seems actively counter-intuitive on Linux.
Developers of Drobo-utils have not tested HFS. Linux does not write to Journalled HFS+
at this point, so HFS support is not present. Good success is reported with Ntfs-3g,
but the performance is much lower than what is typically reported with ext3.
Unless physical movement of the disk to between systems is required, the
native (ext3) format is the best option.
This is a procedure for configuring a Drobo Pro for access via iSCSI. This
information is based on a post by Lemonizer on the Google Group 2009/05/16, with
updates based on improvements and tests by others in the fall of 2009:
1. Connect the Pro via USB, and manually configure the ip of the dbpro
# drobom info settings
# drobom set IPAddress 192.168.2.80
# drobom set NetMask 255.255.255.0
# drobom set UseStaticIPAddress True
The next step is to disconnect USB, and connect by iSCSI:
2. Install open-iscsi (http://www.open-iscsi.org/): sudo apt-get install open-iscsi
3. Connect the dbpro to host machine via iscsi
4. Query dbpro's id: sudo iscsiadm --mode discovery --type sendtargets --portal 192.168.2.80
5. Copy the id string returned by iscsiadm, something like "iqn.2005-06.com.datarobotics:drobopro.tdb091840080.node0"
6. Connect to the dbpro: sudo iscsiadm --mode node --targetname iqn.2005-06.com.datarobotics:drobopro.tdb091840080.node0 --portal 192.168.2.80:3260 --login
If everything went well, your drobopro should show up under /dev. Also check /var/log/messages to
confirm that the iscsi device connected successfully. After that, drobo-utils should be able
to detect the Drobo and manage it over ethernet/iSCSI.
(source: http://groups.google.com/group/drobo-talk/browse_frm/thread/453e02e105e9b41?hl=en )
Some people reported data corruption. This link claims to fix one such
issue:
http://www.drobospace.com/forum/thread/13951/Dropped-iSCSI-connections/?page=2#24792
Upgrading firmware is pretty self-explanatory in the GUI. the first time you
press the Update button, it checks to see if a new firmware is available. If
it there is newer firmware, it offer to upgrade, with suitable prompts.
Similarly, the line mode interface has two commands to deal with firmware,
fwcheck will tell you if an upgrade is required. the fwupgrade
will do the job. It takes a few minutes, and prints a status
you you can see how it is progressing. Have patience:
root@pepino:/home/peter/drobo/drobo-utils/trunk# drobom fwupgrade
validateFirmware start...
Magic number validated. Good.
484 + 2937552 = 2938036 length validated. Good.
CRC from header: 4260378881, calculated using python zlib crc32: 398201869
yeah, the header CRCs do not match. For now they never do ... ignoring it.
CRC for body from header: 1852877921, calculated: 1852877921
32 bit Cyclic Redundancy Check correct. Good.
validateFirmware successful...
writeFirmware: i=484, start=484, last=2938036 fw length= 488
.
wrote 32768 bytes... total: 33252
wrote 32768 bytes... total: 66020
.
.
.
wrote 32768 bytes... total: 2720228
wrote 32768 bytes... total: 2752996
wrote 32768 bytes... total: 2785764
wrote 32768 bytes... total: 2818532
wrote 32768 bytes... total: 2851300
wrote 32768 bytes... total: 2884068
wrote 32768 bytes... total: 2916836
wrote 21200 bytes... total: 2938036
writeFirmware Done. i=2938036, len=2938036
root@pepino:/home/peter/drobo/drobo-utils/trunk#
when it's done, check if it worked using:
root@pepino# drobom status
/dev/sdf - 00% full - (['New firmware installed'], 0)
If the status is like that, then do:
root@pepino:/home/peter/drobo/drobo-utils/trunk# drobom shutdown
lights will flash etc... wait until Drobo goes dark.
Wait another five seconds, then un-plug the USB / connector.
Plug it back in, and wait 10 seconds.
it should start up with the latest firmware available for the drobo.
The drobom commands, like DRI's dashboard, will normally
get the latest and greatest firmware and upgrade. If necessary
one can load arbitrary firmware from the CLI with fwload command.
Those worried about safety of using this software should know: it was
developed with assistance from the vendor (Data Robotics Inc.), and
in every case, based on vendor documentation, and with at least encouragement,
if not outright support. For each release, a QA.txt file is built, demonstrating the functionality tests run. There are multiple checksum verifications built
into the firmware upgrade process, so it is next to impossible to brick a drobo
using the tools. Drobo-utils verifies firmware checksums before attempting
to upload the image to the device, and the device checks the firmware against
the checksums as well. New firmware is loaded into an alternate location
from the currently active one, and if activation of the new firmware fails,
the drobo will simply boot the old one.
On the other hand, common sense rules do apply. Setting the LUN size, or
re-formatting a Drobo will erase all data whether it is done on Linux or
any other operating system. These are power tools, and they can do some
damage data if used without proper care. For example, the reliability
of any storage unit does not reduce the need for backups it only makes doing them
easier. A Drobo is an excellent place to put backups, but not a substitute for
them. Backups are the only way to address error 18 (the number of inches in
front of the keyboard the source of the issue lies.) and no storage unit can
protect against fire or flood.
Drobo has been tested with every old firmware version. Any Drobo should
be upgradable to modern firmware using the dashboard.
for Drobo v1's (only models available to me used for QA)
- NOTE: really need at least 1.1.1 to use Linux & ext3.
- just use the tools to upgrade firmware ASAP.
- 1.01 - very old... bad idea to install this, need to write
- a script to get out, because it isn't in the revision
table. not much works except firmware upgrade.
DO NOT USE. UPGRADE ASAP
- 1.0.2 - works ok in CLI And GUI to view, and upgrade firmware.
- DO NOT USE. UPGRADE ASAP
- 1.0.3 - GUI and CLI work OK, can upgrade firmware.
- Used ntfs3g for a few months under ubuntu 7.10 Linux.
Used 2 TB LUN, with 1.5 TB of physical space available.
DO NOT USE. UPGRADE ASAP
- 1.1.0 - dashboard works no issues.
- from this point, you don't seem to need to unplug the USB
connector to complete the upgrade.
- firmware prior to here deals badly with ext3.
- 1.1.1 - 1.2.4 works without issues.
- ('name' not supported by firmware)
1.3.0 - works without issues.
1.3.5 - works without issues.
No storage unit ever constructed, at any price point, can live upto the expectation
of never losing data. There is no magic wand to wave to solve the data security
problem. People still need a strategy around backups and their maintenance.
Drobo makes it easier to implement a strategy, but does not replace it.
Data Robotics has a best practices page that says it well, but the phrasing
is a bit enterprisy, and while it provides general concepts, it is not
prescriptive enough for people to easily apply.
This section gives some examples & use cases to help people
develop the appropriate strategy for them. Try to keep it simple & concrete.
Don't rely on a single device, ever! Before deploying a storage unit,
one should perform the thought experiment of what will happen if all data on
it is lost. There are always levels of risk. For personal use, one might
accept the risk that if the house burns down, only have infrequent offsite
backups are available and months or years of data may be lost. If someone
is running their business out of their home, this risk will likely not be
acceptable.
The simplest method of backing up your data is to put it in a humungous
single place, and backup the whole thing. That is a valid strategy, but
consider the following realistic case:
A company does incremental backups[#incrbkup] once a day, and full backup
once a week. Once a month, a second full backup is kept as monthly, while the
weeklies are recycled. Monthly backups are kept for a year, and each
year, one backup is kept for five years. So if you write data once and
keep it unchanged for five years, you will have 3 weekly backups, 11 monthlies,
and 4 yearly backups of that data, or 17 copies. This strategy is not
unusual or particularly excessive, many corporate policies end up with 50
or more copies of the data.
With that in mind, if you avoid backing up what you don´t need to, then you
are not saving just one byte, but all the copies too. With a little thought,
one can usually reduce the total storage needs by classifying data appropriately.
Different Drobo models have very different performance. Deploying a Drobo
using a USB interface as primary storage is likely to disappoint.
See the compatibility matrix for details.
Classify Your Data
There are different levels of value for data. Things that are downloaded
from the internet, or source code that is pushed to a repository on the internet,
have natural backups in most cases. The loss of data being queued for printing,
might not be a terrible loss. The loss of videos recorded off the air from
television, might not be terrible either.
Any kind of data which is either not worth backing up, or for which a backups
already exist, does not need to be backed up locally. The other end of the
spectrum is proprietary data, for which copies on the internet are not be
appropriate, and which is irreplaceable if lost. In a photographic business,
the photos, Tax records, accounts, etc... In a personal realm, these would
include family photos, etc... lets call this sort of information precious data
That irreplaceable data is what you need to safeguard. So the classification
can be done in a number of ways, but the simplest is just to only put precious
data in the home directory. So far that´s normal. The unusual thing comes next: Do not put anything else there. Internet downloads, easily replicable data, etc...
should go somewhere else. In general, keep the home directories of users as small
and precious as possible.
As another example, in my personal use case, Linux is readily downloadable, so
there are no system backups at all. Configurations are relatively straightforward,
only credentials, and special configurations are backed up, by having copies
of the information in an normal /home directory of a user. The restoral time
for a single system is not an issue for my personal use, and by the time a
restore is necessary, there will probably a new OS version to try out,
so the value of system backups is quite limited.
The only thing backed up, is the personal (/home) directories of a few users.
If a Drobo is used for primary storage for precious data, a second one should be
obtained as a backup device. It´s as simple as that. Keeping all data on
one device that cannot be backed up is asking for trouble.
Zealots will say that the second unit should be off-site.
The Gen 1 / Gen 2 Drobo´s are also have limited performance, and are perhaps not
well suited to a role as primary storage.
If the data there is all space that either exists elsewhere (mirrors of internet
sites), can be regenerated (object files of compilation, recordings from on-air
broadcasts in a media server), none of this data is particularly precious, and
all of it can be recovered over time in the event of a data loss on one unit.
Again, one argument against such usage is performance. First and second generation
units are a bit on the slow side for use in say. On the other hand, media
serving is a streaming application with typically low instantaneous bandwidth
requirements, so even first generation Drobos should be fine for that.
A rational means of configuring the Drobo is as a backup repository.
The drobo is destination of the backups. Primary copies are on the
desktops & laptops being backed up. Viewed in that way, backup data
is easily recovered in the event of a data loss: just backup the system
again. Of course the history of backups is lost, but the important
thing is usually being able to recover current data.
To find Drobo on a system, drobo-utils queries all the attached devices for indications
it is made by Data Robotics. These strings change from product to product.
If the (new model) Drobo is not detected, then run the command line interface
with the hardware detection debugging output turned out. like so:
# drobom -v 16 status
examining: /dev/sda
id: (0, 0, 0, 0, 'ATA ')
rejected: vendor is ATA (not from DRI)
examining: /dev/sdb
id: (2, 0, 0, 0, 'ATA ')
rejected: vendor is ATA (not from DRI)
examining: /dev/sdc
id: (8, 0, 0, 0, 'Drobo ')
rejected: vendor is Drobo (not from DRI)
returning list: []
No Drobos discovered
Here you see that the vendor string is 'Drobo' which was not a known vendor string
at the time this example was run. so then try:
# drobom -s Drobo status
In other words, take the unknown vendor string and feed it as -s option to tweak
detection of drobom. Your drobo will likely then be picked up.
LUN is an abbreviation of 'Logical UNit'. The origin of the term is
SCSI[#SCSI]_ terminology.
When RAID units became too large for support in the past, and were sub-divided
to present smaller units the operating system. The default LUNSIZE on Drobos
is 2 TiB (adjustable using the tools.) If more disk space (after allowing for
parity/redundancy) than LUNSIZE is installed in a unit, Drobo will show a
second (or even third) LUN. Each LUN shows up in Linux as a separate disk
(examples if the first LUN shows up as /dev/sde, the next will be /dev/sdf,
then /dev/sdg.)
If you think you should see multiple LUNS and you don't, you might have a look at
some kernel settings: make sure that scsi_mod kernel module is loaded, make
sure /sys/module/scsi_mod/parameters/max_luns is > 1.
Drobom view will start up one GUI per drobo, regardless of the number
of LUNS. If asked to format, all LUNS for the device will be formatted.
DRI intends Drobospace.com for owners to talk with one another, except no non-owner
can see the discussions, and early on, there was a lot of input from DRI staff, so
it looked a lot like a support forum, but it really isn't. A lot of owners
objected to these forums being private, so a google group was started for people
to talk with one another, and the discussions to remain public:
http://groups.google.com/group/drobo-talk/topics?hl=en
There is still a great role for Drobo space, in that Tier 3 support analysts
(essentially developers.) sometimes look over there. For tier 3 support, one
cannot expect guaranteed response time, but one may be able to provide some
input into future products or firmware features. It turns out that the
Drobospace forums aren't really for support. but don't take my word for it,
here is above was DRI's take (verbatim from a post on
drobospace by MarkF 2008/08/29) on things:
To contact Data Robotics Inc. for support your options are:
- phone support - technical issues: 1-866-426-4280, Mon-Fri from 8am-5pm PST, excluding Holidays.
- phone support - presales questions: 1-866-99ROBOT
- email support - technical issues: support@drobo.com
- email support - presales questions: sales@drobo.com
- web-based support request: http://www.drobo.com/Support/Request_Support.html
All technical support calls, emails, and web requests are assigned a case number
and tracked. DRI has 3 tiers of customer support. Tier 1 and 2 handle the majority
of cases. They are responsible for tracking phone, email, an web cases and
resolving them. Some cases are escalated to Tier 3 whose personnel reside
in our corporate headquarters and have access to engineering staff.
Support on Drobospace.com:
Drobospace.com is a user community and relies on the volunteer efforts of its members to help each other. Because it is run by volunteers response to problems varies. Tier 3 support personel monitor the Drobospace forums -- tiers 1 and 2 are focused directly on customers, and they are not required to read drobospace.com. By design Tier 3 personnel do not immediately respond to each posting in order to allow the community to function. Depending on the nature of the problem, tier 3 may post in the forum or contact the member directly through a private message to facilitate problem resolution.
Because Drobospace is owned and run by a third party, Capable Networks LLC, there is no linkage with DRI's database systems. Posts here are not assigned case numbers and tracked - that only happens with cases opened directly with DRI.
2 Terabytes is the biggest you should use for now. There are lots
of experiments on the google groups, summarized here: LUNSIZE fits all
Also consult the compatibility matrix indicates best guesses at the
current state of affairs. DRI announced new firmware 1.1.4 for Drobo PRO
which is supposed to remove the 2 TiB restriction, but that isn't confirmed
yet.
That's actually a bit nasty. Nothing happens at first, everything seems
to work fine. After a while, it fails to reclaim space when files are deleted.
The blue capacity lights don't show much relation to how full the file system
is, as reported by the operating system. Drobo may become insatiable, always
asking for more and more disk space, even though the amount of data used
on the file system doesn't warrant it. In extreme cases, data may become
in-accessible.
DRI naturally releases new versions of firmware and may fix these issues
at some point. If you are willing to test it out on your new Drobo, the
procedure to do so is simple:
1. Create a file system as per normal.
2. Fill the physical space up. (blue lights should light up.)
3. Remove alot of the data.
If the problem is not there, then blue lights will function properly and go
out to correspond to the deleted data after a while. If the blue lights
do not go out after step 3, then do not trust your data to this file system.
Re-create with a smaller LUN, and try again. 2 Terabytes is the only case
of documented success so far.
Short answer: no.
For Drobo to perform storage management, it has to know what space is free, so it
needs to understand the file system you are using. The list of file systems
it official understands is: FAT32, NTFS, HFS+, ext3. That's it, so if you want
to use reiserfs, or xfs, or GFS, or whatever... you are doing research. The
vendors says those other file systems types will not work.
No. The way the data is placed on the disks is completely proprietary.
You cannot take the drives and connect them individually to a server, and read
the data off that way, because it isn't a linux md or lvm format that can easily
be reconstructed. You cannot give the disk pack to a data recovery company,
because they do not know the data format either, and you will have to pay them
to reverse engineer DRI's format, which will get expensive quickly.
You need either a backup, or another Drobo. Even among Drobos there
are limits to compatibility see the Drobo.com web site for details.
The order and timing of disks being connected to hot-plug busses will
determine the device name (it might be /dev/sdb one time, and /dev/sdc another.)
So putting /dev/sdX in the fstab to mount their disks, as is traditionally done,
won't work. Instead, do:
peter@pepino:~$ ls -l /dev/disk/by-uuid
total 0
lrwxrwxrwx 1 root root 10 2009-12-15 04:54 2C88743C8874071C -> ../../sda3
lrwxrwxrwx 1 root root 10 2009-12-15 04:54 32a41d0a-b193-41f3-86fa-29bbee8cd2b3 -> ../../sda8
lrwxrwxrwx 1 root root 10 2009-12-26 12:08 3cd5d9cc-c227-4ed8-bab2-60c2d71f6e9d -> ../../sdf1
lrwxrwxrwx 1 root root 10 2009-12-15 04:54 72b0ee8c-d0e8-479d-b79c-3dbda1581f55 -> ../../sda6
lrwxrwxrwx 1 root root 10 2009-12-15 04:54 814472db-dbee-411c-8870-7ca59f32e7c1 -> ../../sda5
lrwxrwxrwx 1 root root 10 2009-12-26 12:16 8ed93296-9be2-4576-9ae4-9d9c78363fb6 -> ../../sdg1
lrwxrwxrwx 1 root root 10 2009-12-15 04:54 a4bc252e-0eb7-489c-94e7-688efd528665 -> ../../sda7
lrwxrwxrwx 1 root root 10 2009-12-15 04:54 bc1ab400-df49-457d-8700-c77dde19e450 -> ../../sda2
lrwxrwxrwx 1 root root 10 2009-12-15 04:54 C2EE700DEE6FF7D5 -> ../../sda1
peter@pepino:~$
The UUID is a name that is constant for a partition. Each time a partition
is mounted, a link will be created in this directory towards the correct
/dev/sdX. A UUID related /etc/fstab entry looks like:
UUID=3cd5d9cc-c227-4ed8-bab2-60c2d71f6e9d /drobo01 ext3 defaults 0 2
because an ext3 file systems is an ext2 file system with a journal.
The normal command to build an ext3 file system is mke2fs -j.
DRI: Multiple partitions per LUN is supported. Having any combination of file
supported file systems on the different LUNS and partitions is fine as well.
Some people do it. I would not risk it.
The Linux Volume Manager is a layer of software which is shimmed between the file system layer, and the physical disks. It provides a 'fake' (virtual) volume on which file systems are built. This gives flexibility to concatenate several physical volumes together to make a single file system, or allocate a single volume to different file systems over time, as needs dictate rather than all at the outset.
For Drobo, LVM would be especially cool in that one could initially allocate only
the physical space actually available within the LUN, and thus applications which
key on avoiding filling file systems would function correctly, instead of always
asking to insert more storage, and not managing the storage available. When more
physical space (new drives!) becomes available, one could allocate more space to
the virtual volume, and then grow the file system. So Drobo would still take
care of the drudgery of RAID set maintenance, relay operations, etc... but the
user would have more control on how space was allocated at the OS level.
For that to work, Drobo firmware would have to understand LVM to some extent.
Initial experiments showed that the firmware didn't appreciate LVM. Being
able to add a Drobo unit as an pv to an LVM unit would be very cool.
Drobo works by knowing how the file system is laid out and pretending to have
more space than is physically present. Drobo does some sleight of hand to
manage disk space and move things around optimally as hard disks fail or are
added.
Full disk encryption implemented by the operating system makes the Drobo unable
to understand the file system, so it doesn't know which blocks are in use, and
the unit will always believe the system is completely full. Drobo will not
behave well. Among the methods that will not work are any that operate on a
raw disk partition, such as truecrypt, or any of the linux cryptoloop based solutions.
Instead of whole disk encryption, a method that uses an underlying file system
that is well known to the Drobo (the list is short: FAT32, NTFS, HFS, EXT2) is
needed. On windows, encrypting directories with standard NTFS will work fine.
On Linux, a good choice would be EncFS http://www.arg0.net/encfs, which encrypts
file names and data over an ext file system, or some other method which uses
FUSE http://fuse.sourceforge.net. is reported to work well.
Best first stop is the google Group.
General Public License - Gnu - GPL .
who did what:
Peter Silva: wrote most all of it.
Chris Atlee: the proper debian packaging.
Brad Guillory: some help with diagnostics and patches.
Joe Krahn: lots of inspiration.
Andy Grover: some elegance cleanups.
Sebastian Sobolewski: DroboPro patches, and testing.
Testers (of DroboPRO):
robj, Sebastian (aka Tom Green), ElliotA, Andrew Chalaturnyk
version 9999, somewhen
copyright:
Drobo Utils Copyright (C) 2008,2009 Peter Silva (Peter.A.Silva@gmail.com)
Drobo Utils comes with ABSOLUTELY NO WARRANTY; For details type see the file
named COPYING in the root of the source directory tree.
This software is copyright under GPL. See near end file for details...
This is a pile of random useful bits for developers. (aka notes to myself ;-)
Project Source code is managed using a git repository provided by
sourceforge.net. Git ( http://git-scm.com/ ) provides a fully
distributed development model, so one can exchange patches arbitrarily
among developers. On the sf.net repository, the 'master' branch is
the final integration target for future releases.
the following checks out the master branch of the source code tree and
puts it in the drobo-utils subdirectory of the current tree.
git clone ssh://username@drobo-utils.git.sourceforge.net/gitroot/drobo-utils/drobo-utils drobo-utils
Before doing Any commits, ensure that the author fields are
appropriately set. One can obtain patches applied to the branch
with git pull, and commit patches for distribution with git push.
To build the package from source, there are a number of other packages needed:
boule% sudo aptitude install debhelper python2.5-dev
boule% sudo aptitude install python-docutils
boule%
The second line only required to build documentation. On the other hand,
to get a complete list of packages you need, it is best to use a shell window
to grep in the Debian package control file (which defines what the
dependencies are for the build system):
peter@pepino% grep Depend debian/control
Build-Depends: debhelper (>= 5), python2.5-dev, python-docutils
Depends: ${shlibs:Depends}, ${misc:Depends}, parted
peter@pepino%
Sometimes, when there are issues, the correction gets checked in, but there
hasn't been time to do a full release process. If you really need the fix,
then your only option is to get it from the developers' git repository.
If you are told 'it is in git', that refers to the git source code management
system, and the server for that is at sourceforge.net. How to get it:
# git clone ssh://developer@drobo-utils.git.sourceforge.net/gitroot/drobo-utils/drobo-utils mine
# cd mine # go into the source directory you downloaded
# ./drobom status # try it out...
# python setup.py install # install it in system places...
# drobom status # try it out...
# git pull # get any changes made since the git clone was done.
# python setup.py install # install the changes in system places.
Sample checkout of a stable version. To view available branches:
% git branch -r
origin/HEAD -> origin/master
origin/master
origin/noC
origin/peter
origin/r0.2.1
origin/r0_3_3
origin/r0_4_0
origin/r0_5_0
origin/r0_6_0
Where a version is something like r0_4_0. then you can pick anyone to work with:
% git branch r0_4_0 -r origin/r0_4_0
Branch r0_4_0 set up to track remote branch r0_4_0 from origin.
% git checkout r0_4_0
Switched to branch 'r0_4_0'
%
When you use git to get a tree, it keeps copies of metadata to be able to
track changes. If you want a copy that is contains no git cruft,
rm -rf .git in the root of the source tree.
If you are mixing downloaded packages and source installs, check out the next
section for gotchas.
The 'setup.py' script, mentioned in the previous section, is a convention
from the distutils python packaging system. distutils installation is slightly
different from installation from debian packages. There doesn't seem to be a
distutils way to remove a package. touch all the files, do an installation, then
manually remove the files it installed.
drobo-utils has been picked up for inclusion in debian. The "real" packaging
for debian packages is kept in a separate tree, and maintained by debian
developers.
The debian/ setup puts stuff in /usr/sbin while setup.py puts things in /usr/bin. Python install does not install man pages either, which the dpkg takes care of. The libs are placed differently too. haven't reviewed for other conflicts, least confusing to use one or the other method on a system.
(if you do distutils install, then remove the debian package via: dpkg --purge drobo-utils)
Procedure:
1 - make a branch
# assuming you have a local repository...
git branch <branch> # creates the branch, from the cwd (ought to be master)
git checkout <branch> # switches current dir to the branch.
2 - Stamp the branch with version
vi CHANGES.txt # complete change manifest for release
vi debian/changelog # copy manifest from txt, add signature.
vi setup.py # edit version
vi Drobo.py # edit VERSION
3 - Build packages for testing & Install them. (see separate recipe.)
4 - Run QA.
Record results of release tests in the branch (QA.txt) As new tests are created,
modify QA.txt on trunk for to keep references for the next release.
5 - Commit & Push QA'd branch
git commit -a
git push origin origin:refs/heads/<branch>
QA.txt is a quality assurance log. The version on the trunk of the releases
indicates the QA procedure to be applied to the next version during the
release process. Since a branch is created for each release, the version
of QA.txt acts as a quality log for that release. so one can do a git
checkout, or git export to get the quality log for any release (QA.txt
introduce in version 0.4.0)
Assumes you have installed the Build dependencies:
# obtain a fresh tree
% git clone ssh://user@drobo-utils.git.sourceforge.net/gitroot/drobo-utils/<version> drobo-utils-<version>
% cd drobo-utils-<version>
% rm -rf .git # get rid of Git cruft, yielding a raw source tree.
% chmod 755 debian/rules # I dunno why the permissions are wrong...
# this debian/ config is just for non-distro packages.
# builds for debian and Ubuntu.
% dpkg-buildpackage -rfakeroot
% cd ..
# rename it for whatever distro is appropriate...
% mv drobo_utils_0.3.3-1_i386 --> droboutils_0.3.3-1_i386_ubunutuIntrepid.deb
# rebuild the source tar because it will have the 'debian' link in it.
% cd drobo-utils-0.99.9
% rm debian
% cd ..
% tar -czvf drobo-utils-0.3.3-1.tgz drobo-utils-0.99.9
apply QA tests. as per QA.txt recording results there.
use the restructured text tools (from the python-docutils package.)
to build things using:
% make doc
Have a look at Makefile for how that works.
update the web site:
% scp README.html <user>,drobo-utils@web.sourceforge.net:htdocs
Droboshare is not directly supported by drobo utils running on a linux host.
However, the droboshare itself is a linux host, and it is possible to run
drobo-utils un-modified on the droboshare itself. There is download called
the Droboshare Augmented Root File system (DARFS), which includes a python
interpreter and drobo-utils.
Why isn't there full support in host based drobo-utils itself? Digital
Robotics hasn't released details of the protocol used by the proprietary
dashboard to communicate with a droboshare, so it would be a lot of work to
reverse engineer that. So support of a droboshare from a linux GUI on a
host system is not likely in the near future.
If someone wants to figure that out, it might be a good thing (tm)
After that is figured out, the next step would be to understand
how to flash the firmware remotely. That would eliminate the last function
that cannot be done with open source.
Drobo is the only consumer-level storage unit that does Thin Provisioning (allocating a device larger than the physical space available, allowing space upgrades without
OS changes.) Drobo does this by understanding the file system blocks, which is
why it only supports a very limited set of file systems and cannot support full
disk encryption.
There is considerable industry activity about adding ATA TRIM and corresponding
SCSI UNMAP commands. These commands, for their respective command sets, add
the ability for the operating systems' file system code to indicate blocks that
are not in use to storage units. Drobo would work with any file system that
uses these commands, with far less firmware. On linux, that file systems that
are starting to support TRIM/DISCARD are: ext4, btrfs, and xfs. It may also
help with the inherent limitations around full disk encryption.
These commands are still maturing in support. Long term, they seem like
The right thing to do.
(2009/12/30)
Revision date: 2009/12/27
copyright:
Drobo Utils Copyright (C) 2008,2009 Peter Silva (Peter.A.Silva@gmail.com)
Drobo Utils comes with ABSOLUTELY NO WARRANTY; For details type see the file
named COPYING in the root of the source directory tree.
version 9999, somewhen
[Links and Help][ Index]
Droboshare is not directly supported by drobo utils running on a linux host.
However, the droboshare itself is a linux host, and it is possible to run
drobo-utils un-modified on the droboshare itself. A python interpreter is needed
to run drobo-utils. A python interpreter has, itself, a number of dependencies.
So you number of packages need to be installed on the droboshare.
This is where DARFS comes in.
The Droboshare Augmented Root File System (darfs) is a 60 MB or so download
you can get from drobo-utils.sf.net. There isn't any source code, because,
well, nothing from any of the packages has been modified. there are
instructions on how to build DARFS in DEVELOPERS
DARFS is a standard droboshare root file system, with some packages added:
openssl, openssh, berkeleydb, bzip2, a fairly complete Python 2.6.2. drobo-
utils is a python app. and it works in line and API mode, natively, on
the droboshare. for example, I've used it to replace the firmware. no
problem at all.
People un-afraid of the command line can upgrade drobo firmware, query
status, and take diagnositc dumps, from the command line on the droboshare
itself, just as they would on any linux host computer. But a full GUI
would be too much for the little processor and more importantly the limited
memory in the droboshare, so that is not provided.
Download it from drobo-utils.sf.net:
steps:
- copy the tar file onto somewhere on your share.
- log in via DropBear ssh as a root user on the droboshare.
- cd /mnt/DroboShares/YourDrobo (root of drobo file system, for example)
- tar -xzf darfs.tgz (root of drobo file system, for example)
- the root directory of the tar is 'slash'.. it will be under YourDrobo
- export PATH="/mnt/DroboShares/YourDrobo/slash/usr/bin:${PATH}" (which is where python and drobom are.)
- drobom status
you're done!
all you need to do is:
ln -s /usr/libexec /mnt/Droboshares/YourDrobo/slash/usr/libexec
Try an sftp from another machine (as root...) and it ought to work.
(explanation: when one tries to sftp to a droboshare, it gives an error
about trying to exec '/usr/libexec/sftp-server'. Openssh builds the
right binary, but Dropbear doesn't know where to look for it. the
libexec directory isn't there on the droboshare, so there is no harm
in creating it and linking into DARFS.)
With DARFS, and the third party software you can get from drobospace and
drobo.com, the droboshare is very open and hackable. However, there
remains a one limitation: There is no open source way to upgrade or modify
droboshare firmware. If you want to re-flash to a factory original
state, you need the vendor dashboard.
DARFS was build in late 2008 as a technology demonstration. There was not
much interest at the time, so it has not been followed up since that time.
If there is sufficient interest, may take it up again. The quality of the
build documentation procedure is, well, poor. It would have been necessary
to perform the entire procedure from scratch to verify the details. The
information here is basically notes to the author. YMMV.
I used this procedure on ubuntu jaunty, but the same packages ought to work
identically on Debian Lenny. the idea here is to use the fabulous scratchbox2
tools developed for use with the Nokia Maemo.
Cross compiling is usually a PITA (excuse me.) you want to use a nice fast,
powerful desktop machine (the "host") to compile something that will run on a
much less muscular embedded systems (the "target") Scratchbox 2 looks at a
binary, and if it is for the target cpu to run, then it invokes the qemu
processor emulator so that it will run on the host platform under emulation.
When the binary found is a host binary, it just runs it normally. This makes
cross-compilation very transparent. The other magick that it does is to
invoke the cross-compiler when the package's normal build procedure will try
to invoke a native compiler. Scratchbox removes a great deal of pain from
cross-compiling.
So the steps that follow will allow you to build an environment where it is
easy to build additional software as desired. It is worth noting that the
droboshare has a cpu that is relatively powerful, and relatively idle in
most cases. It is only prevented from being really useful by the limite
memory available.
You need a droboshare, with a drobo on it.
You need also need to download some stuff:
mkdir ~/drobo
cd ~/drobo
mkdir droboshare
cd droboshare
mkdir Downloads
cd Downloads
Drobo SDK: http://www.drobospace.com/download/11742/Data-Robotics-SDK/
You need to get root access via Drobbear, read the SDK documentation for that.
# which ought to get you ... SDK.zip
cd ..
mkdir sdk
cd sdk
unzip ../SDK.zip
You also need (as per the SDK) a code sourcery toolchain: http://www.codesourcery.com/sgpp/lite/arm/releases/2006q1-6 with target platform: ARM Gnu/Linux
sudo apt-get install scratchbox2... it might look like this:
trestler:/etc/apt# apt-get install scratchbox2
Reading package lists... Done
Building dependency tree
Reading state information... Done
The following packages were automatically installed and are no longer required:
libisc44
Use 'apt-get autoremove' to remove them.
The following extra packages will be installed:
bochsbios debootstrap libbrlapi0.5 libicu38 libvdemgmt0 libvdeplug2 openbios-sparc
openhackware proll qemu vde2 vgabios
Suggested packages:
sudo samba sbrsh vde2-cryptcab kvm
The following NEW packages will be installed:
bochsbios debootstrap libbrlapi0.5 libicu38 libvdemgmt0 libvdeplug2 openbios-sparc
openhackware proll qemu scratchbox2 vde2 vgabios
0 upgraded, 13 newly installed, 0 to remove and 0 not upgraded.
Need to get 17.7MB of archives.
After this operation, 49.2MB of additional disk space will be used.
Do you want to continue [Y/n]?
Get:1 http://gulus.usherbrooke.ca lenny/main libicu38 3.8.1-3 [5918kB]
Get:2 http://gulus.usherbrooke.ca lenny/main libvdemgmt0 2.2.2-3 [12.4kB]
Get:3 http://gulus.usherbrooke.ca lenny/main libvdeplug2 2.2.2-3 [11.6kB]
Get:4 http://gulus.usherbrooke.ca lenny/main openhackware 0.4.1-4 [76.4kB]
Get:5 http://gulus.usherbrooke.ca lenny/main proll 18-4 [248kB]
Get:6 http://gulus.usherbrooke.ca lenny/main libbrlapi0.5 3.10~r3724-1+b1 [62.7kB]
Get:7 http://gulus.usherbrooke.ca lenny/main vgabios 0.6b-1 [79.1kB]
Get:8 http://gulus.usherbrooke.ca lenny/main bochsbios 2.3.7-1 [155kB]
Get:9 http://gulus.usherbrooke.ca lenny/main openbios-sparc 1.0~alpha2+20080106-2 [229kB]
Get:10 http://gulus.usherbrooke.ca lenny/main qemu 0.9.1-10 [10.5MB]
Get:11 http://gulus.usherbrooke.ca lenny/main vde2 2.2.2-3 [181kB]
Get:12 http://gulus.usherbrooke.ca lenny/main debootstrap 1.0.10lenny1 [52.1kB]
Get:13 http://gulus.usherbrooke.ca lenny/main scratchbox2 1.99.0.24-2 [150kB]
Fetched 17.7MB in 34s (518kB/s)
Selecting previously deselected package libicu38.
(Reading database ... 54086 files and directories currently installed.)
Unpacking libicu38 (from .../libicu38_3.8.1-3_i386.deb) ...
Selecting previously deselected package libvdemgmt0.
Unpacking libvdemgmt0 (from .../libvdemgmt0_2.2.2-3_i386.deb) ...
Selecting previously deselected package libvdeplug2.
Unpacking libvdeplug2 (from .../libvdeplug2_2.2.2-3_i386.deb) ...
Selecting previously deselected package openhackware.
Unpacking openhackware (from .../openhackware_0.4.1-4_all.deb) ...
Selecting previously deselected package proll.
Unpacking proll (from .../archives/proll_18-4_all.deb) ...
Selecting previously deselected package libbrlapi0.5.
Unpacking libbrlapi0.5 (from .../libbrlapi0.5_3.10~r3724-1+b1_i386.deb) ...
Selecting previously deselected package vgabios.
Unpacking vgabios (from .../vgabios_0.6b-1_all.deb) ...
Selecting previously deselected package bochsbios.
Unpacking bochsbios (from .../bochsbios_2.3.7-1_all.deb) ...
Selecting previously deselected package openbios-sparc.
Unpacking openbios-sparc (from .../openbios-sparc_1.0~alpha2+20080106-2_all.deb) ...
Selecting previously deselected package qemu.
Unpacking qemu (from .../qemu_0.9.1-10_i386.deb) ...
Selecting previously deselected package vde2.
Unpacking vde2 (from .../archives/vde2_2.2.2-3_i386.deb) ...
Selecting previously deselected package debootstrap.
Unpacking debootstrap (from .../debootstrap_1.0.10lenny1_all.deb) ...
Selecting previously deselected package scratchbox2.
Unpacking scratchbox2 (from .../scratchbox2_1.99.0.24-2_i386.deb) ...
Processing triggers for man-db ...
Setting up libicu38 (3.8.1-3) ...
Setting up libvdemgmt0 (2.2.2-3) ...
Setting up libvdeplug2 (2.2.2-3) ...
Setting up openhackware (0.4.1-4) ...
Setting up proll (18-4) ...
Setting up libbrlapi0.5 (3.10~r3724-1+b1) ...
Setting up vgabios (0.6b-1) ...
Setting up bochsbios (2.3.7-1) ...
Setting up openbios-sparc (1.0~alpha2+20080106-2) ...
Setting up qemu (0.9.1-10) ...
Setting up vde2 (2.2.2-3) ...
Setting up debootstrap (1.0.10lenny1) ...
Setting up scratchbox2 (1.99.0.24-2) ...
trestler:/etc/apt#
The idea here is to get a starting point by making a copy of the root file
system from a droboshare. After downloading the the SDK from drobospace.com,
and enabling root shell access, just log into the droboshare and:
trestler:/etc/apt# ssh root@droboshare
The authenticity of host 'droboshare (172.25.5.13)' can't be established.
RSA key fingerprint is 90:75:3d:ca:f1:42:65:92:71:97:48:d7:6b:ff:d7:8b.
Are you sure you want to continue connecting (yes/no)? yes
Warning: Permanently added 'droboshare,172.25.5.13' (RSA) to the list of known hosts.
root@droboshare's password:
Welcome to Embedded Linux
_ _
| ||_|
| | _ ____ _ _ _ _
| || | _ \| | | |\ \/ /
| || | | | | |_| |/ \
|_||_|_| |_|\____|\_/\_/
A Data Robotics Product.
http://www.drobo.com/
BusyBox v1.1.2 (2007.06.18-15:03+0000) Built-in shell (ash)
Enter 'help' for a list of built-in commands.
~ $cd /
/ $ tar -cvf /mnt/Droboshare/Dro*/droboshare_root.tar bin dev fs.ls lib opt sbin src usr version boot etc home linuxrc root serial tmp var
on your Debian or Ubuntu system,
mkdir slash
cd slash
tar -xvf slash.tar
#in the SDK there is some libz stuff, add it into the appropriate places (/usr/lib for libz, and /usr/include for a .h)
cd slash
cd tmp
tar -xvf ~/drobo/droboshare/sdk/libz.tgz
mv *.so ../usr/lib
mv *.h ../usr/include
cd ../..
tar -xjvf arm-2006q1-6-arm-none-linux-gnueabi-i686-pc-linux-gnu.tar.bz2
creates the ~/drobo/droboshare/armx directory for the toolchaimkdir slash
cd slash
tar -xvf slash.tar
#in the SDK there is some libz stuff, add it into the appropriate places (/usr/lib for libz, and /usr/include for a .h)
cd slash
cd tmp
tar -xvf ~/drobo/droboshare/sdk/libz.tgz
mv *.so ../usr/lib
mv *.h ../usr/include
# add includes from the cross-compilation environment...
# this isn't quite right, the includes came from somewhere else... have to look around...
cd ~/drobo/droboshare/
dirs="`find armx -type d -name include`"
#copy all the include from the sub-directories in armx into slash/usr/include...
#something like this might work:
for i in $dirs; do
cd $i ; tar -cf - . | (cd ~/drobo/droboshare/slash/usr/include; tar -xvf - )
done
#but a lot of them seem to be repeats, the only really interesting one is:
./arm-none-linux-gnueabi/libc/usr/include
Build procedures is basically out of the box after this point:
cd bzip2
make
make install PREFIX=/home/peter/drobo/droboshare/slash/usr
cd ../db*
cd build_unix
export CC=/home/peter/drobo/droboshare/armx/bin/arm-none-linux-gnueabi-gcc
./configure --prefix...
make
make install
cd ../ncur*
./configure --prefix=...
make
make install
cd ../openssl*
./config --prefix=...
make
make install
cd ../Pyth*
./configure --prefix=
make
Failed to find the necessary bits to build these modules:
_sqlite3 _tkinter bsddb185
gdbm readline sunaudiodev
To find the necessary bits, look in setup.py in detect_modules() for the module's name.
Failed to build these modules:
_curses _curses_panel
running build_scripts
make install
cd ../drobo-utils/trunk
/python setup.py install
cd into usr/bin
vi drobom
#!/usr/bin/env python
:wq
cd ~/drobo/droboshare/src
tar -xzvf ../Downloads/libpcap-1.0.0.tar.gz
cd libpcap*
sb2
./configure --prefix=/home/peter/drobo/droboshare/slash/usr
make
make install
exit
cd ..
tar -xzvf ../Downloads/tcpdump-4.0.0.tar.gz
cd tcpdu*
./configure --prefix=/home/peter/drobo/droboshare/slash/usr
make
fails with:
/usr/lib/libcrypto.a(dso_dlfcn.o): In function `dlfcn_bind_func':
dso_dlfcn.c:(.text+0x2d4): undefined reference to `dlsym'
dso_dlfcn.c:(.text+0x344): undefined reference to `dlerror'
/usr/lib/libcrypto.a(dso_dlfcn.o): In function `dlfcn_bind_var':
dso_dlfcn.c:(.text+0x400): undefined reference to `dlsym'
dso_dlfcn.c:(.text+0x46c): undefined reference to `dlerror'
/usr/lib/libcrypto.a(dso_dlfcn.o): In function `dlfcn_unload':
dso_dlfcn.c:(.text+0x4e8): undefined reference to `dlclose'
/usr/lib/libcrypto.a(dso_dlfcn.o): In function `dlfcn_load':
dso_dlfcn.c:(.text+0x56c): undefined reference to `dlopen'
dso_dlfcn.c:(.text+0x5d4): undefined reference to `dlclose'
dso_dlfcn.c:(.text+0x5ec): undefined reference to `dlerror'
collect2: ld returned 1 exit status
make: *** [tcpdump] Error 1
vi Makefile
/-lc
a-ldl
:wq
make
make install
so there you go, samples of building a whole bunch of packages.
copyright:
Drobo Utils Copyright (C) 2008,2009 Peter Silva (Peter.A.Silva@gmail.com)
Drobo Utils comes with ABSOLUTELY NO WARRANTY; For details type see the file
named COPYING in the root of the source directory tree.
[Links and Help][ Index]
Below are highlights included in each release.
TODO:
These are all Open Tasks... HELP WANTED!
add python version testing to test suite (2.4, 2.5, 2.6)
add LUN test procedure to DEVELOPERS.txt
looking at firmware based LUN creation/deletion & format.
disk pack status.
alerts.
fix make_tarball to exclude .git and debian, like it is supposed to.
review getSubPageStatus, DMIP seems better now.
review unitstatus vs. diskpack status.
- make a daemon. reasons:
-- updated for relay out progress.
-- alerting
-- unified JSON interface for GUI & WUI.
-- managed queueing for device (not five GUI's querying in parallel.)
-- preserve option settings across reboot? (one report.)
-- periodically synch. the drobo clock to the server.
- 0.6.2.2: 2010/02/06
- fix for Redhat/python 2.4 wanting to partition everything!
- 0.6.2.1: 2009/12/30
- fix for Drobo S slotinfo firmware bug (reports 8 instead of 5.)
- 0.6.2: 2009/12/26
- GUI Option setting, thresholds, etc...
- changed command line syntax to more consistently set <item>
- documented option setting in CLI
- more input checking to CLI.
- improved LUN info display accuracy.
- debugged DroboPRO support (debugged via google group. Thanks All!)
- added scsi to info printing options in CLI.
- 0.6.1: October 7th, 2009
- DroboPro identification fix by spobole,
- parted fix from Patrick Bills
- Looked at newly released DMIP. Found option2 commands for settings.
- Option query/set implemented (but cannot test for DroboPRO)
- migrated source code managers from subversion to git
- minor: consistency of default location to look for fw between lib and gui.
- fixed that inability to read lun size on old firmwares crashed GUI.
GUI now works all the way back to the beginning.
- 0.6.0
- Major: GUI heavily re-vamped. Behaves better.
- Major: DroboPRO support added (thanks to lemonizer on the google group!)
- Major: GUI gets online documentation.
- GUI: layouts replaced fixed dimensions.
- GUI: added Rename support
- GUI: added Load firmware button
- GUI: Show Diagnostics (equiv. to diagprint.)
- GUI: Name display to main display. Easier to tell which Drobo is which.
- GUI: added mount point display to space used tool tip.
- api: added ability to detect un-documented feature bit settings.
- ddiag.c becomes drobom diagprint
- catlee fixes the packaging again.
- 0.5.0 2009/01/30
- replaced DroboDMP C extension by DroboIOctl python class. no need
for architecture dependent packages. this also means you can just
try it out straight after download, no need to install anything.
much easier. fewer build dependencies as well.
- device detection completely re-written. It works better now.
- Now displays mount points where it makes sense to do so.
- status command output better/different. shows all devices and mount points
for each drobo.
- full Multiple LUN support. should do what is expected.
- full multiple Drobo support. should do what is expected.
- added modular info printing with csv list of outputs.
- firmware 1.3.0, has enabled formerly broken functionality which
name makes the sync command rename every drobo to 'hi there'.
fixed.
- added 'name' command, to name Drobo. (functionality of Sync)
- corrected detection logic, which depended on formerly immutable name.
(affects fw >= 1.3.0) now use same mechanisms as sg_scan from sg3_utils.
- documentation moved into restructured text. now shows up on the web.
web site and documntation now consistent.
- 0.4.0 Release: 2008/12/28
- IMPORTANT: firmware 1.3.0 support added. Earlier drobo-utils
versions will not recognize a drobo running 1.3.0 or later.
- CLI is more normal... use getopts (got rid of device as fixed place arg.)
now allows command to be an option as well.
- Changed DEBUG from a flag requiring code modification, to a bitmap
honored as a command-line argument (--verbose)
misc. improvements in documentation.
- droboview now only launches for first drobo found in list, and stays foreground
always.
drobom list returns a list that is easier to parse (suitable for use in backticks.)
droboview is now just a stub for drobom view. reduced overhead.
now reports VERSION id in usage. There is a -V option too.
- 0.3.3 release: 2008/11/05
- lunsize display bug fixed.
- Chris's man pages & help improvements.
- more messaging fixups for when you don't invoke as root.
- fix for ubuntu Intrepid deciding all disks are Drobos. !!
- fixed all the firmware info parsing issues.
- firmware load issues should all be gone.
- Brad's fix for ddiag.c, done properly this time (my bad!)
0.3.2 release 2008/10/25
for v1, still don't understand index.txt file.
after testing, some more fixes, v2 downloads now work.
- fixes received from Brad Guillory for v2/tdz firmware downloading.
they don't do any harm afaict, don't have a v2 to test with.
added fwload directive to drobom.
added root user check to give a bigger hint.
0.3.1 release 2008/10/01
- OK I know checking for firmware updates doesn't work
right now, but I need to find out why. code seems correct.
there is a bugfix for getting rid of the 'licensed' part.
- fixes for firmware validation:
- header CRC on 32bit was broken on 64bit.
- 32 vs. 64 bit kludge needed.
- 0.3.0 release 2008/09/04
added ability to format Drobo
added ability to set lun sizes
- bugfix for firmware:
- CRC32 on 32bit intel. (use signed, instead of unsigned.)
- fixed header CRC not validating.
added ability to set time on Drobo.
added simulation mode to Drobo.py, development aid.
some work done towards lunsize setting, but incomplete.
some work towards indicating relay out progress, not tested...
- 0.2.2 Released 2008/08/08
-
- 0.2.1 Released: 2008/08/06
- uploaded to drobospace & sourceforge.
- firmware upload works for .tdf's
- 0.1.1 Initial version uploaded april 2008
- to drobospace.com and sf.net
- confirmed to work with firmware 1.1.1
- shows status, disk slots filled, model, etc..
- shows fw version loaded.
- blink & standby commands work.