Hello Jaime,
I think you bring up some very good ideas about e-smith. I agree it has a
lot of potential. I also agree that a journaled filesystem as well as
additional drives for backups would be a benefit. But...also believe that
the philosophy that e-smith...err Mitel is following is best, that is to
say they keep everything very simple, and still try to make their software
work on the lowest common denominator.
Case in point: They have stayed away from the 2.4 kernel series. Some
people view that as a lack of initiative. I view it as wise. Sure 2.4 gives
us newer drivers, but mainly all it gets us is support for much larger
number of processors and huge amounts of RAM, etc. This is not the kind of
hardware or market they are targeting, and 2.4 is still not as stable as
2.2 is right now, and using it would break too many other things. By the
same token they have not added ReiserFS of JFS of XFS or ext3. Why? They
have not been proven yet. I know SuSE and others believe that they are
ready for primetime. Heck, my test machine running RedHat 7.2 beta is
formatted with the ext3 filesystem right now, and my BeBox has had a 64-bit
journaled filesystem for many years now. But again, this adds levels of
uncertainty not only for configuration but support as well. In the target
market mentioned above, it is just not ready yet. Don't get me wrong. I
hate having to wait through a fsck as much as the next guy, but no
journaled filesystem is ever a substitute for a good tape backup or backup
to a desktop file. (I do agree that tapes are expensive. Once you go past
20gig, the best value out there is the Ecrix series which goes up to 66gig
compressed and now comes in IDE and SCSI and run about $850). Tapes are
valuable in that your can take them out of the data room and put them in a
safe somewhere. If someone forgets to change the tapes, they are only
shooting themselves in the foot.
The backup topic allows me to segue nicely into the idea of the second
drive for backups. I like that one. I like it a lot. But not every left
over machine that Mitel thinks people may put the software on has two
drives. Drives are cheap you say? True, but again it adds levels of
complexity to setup and support. It forces too many assumptions and guesses
on the part of the developers. This has been discussed in detail on the
list before when questions arose about the default partitioning scheme.
Have a simple /boot and / partition makes it much easier for flexbackup to
work with. Adding other drives are partitions only makes life more
difficult.
My recommendation? Get an IDE RAID card. This solves the software issue
altogether. The 3ware series seems to be a very good value at $129 for the
6200 2-port model, and the products from ARCO (http://www.arcoide.com) seem
to work well too. Get yourself two cheap 30gig IDE drives and you have
yourself a nice RAID1 mirror. You can go SCSI, but that gets us back to the
cost issues again.
As far as a UPS being expensive, I agree as well. I bought a TrippLite
1400R rackmount unit to use with 8 outlets and it was about $600 is I
recall. I was also able to get their PowerAlert software to install and
work on e-smith with some modification. It all comes down to how much is
your system and data worth. No matter what the power is going to go out
sometime. $600.00 starts to not seem like so much.
A lot of the hardware ideas you suggest have already been tried by some
people. Net Integration Technologies has their NetIntegrator box which
works very similar to e-smith. They have a base load of the OS stored on a
flash disk, that can be instantly be transferred to the main hard disk at
the touch of a button on the front panel. They also have an LCD display for
monitoring and so forth. The problem? Again, it locks you in to certain
hardware requirements to make use of the software. You do not have the
freedom to build your own box. Of course, their solution is not open source
either :)
The monitoring add-on you have created is very well done. The combo of that
with the log monitor rpm (or Blade :) is a tremendous help to remote
administer a box. The benefits of these utilities is that they can be used
by any type of system they are installed on, regardless of the physical
configuration. They are also based on proven programs/technology. They
followed the simple architecture rule that e-smith started with.
So what's my point? Keep it simple, keep it so that any change that is made
to the distro makes sense from a stability standpoint, a functionality
standpoint, and from the perspective that most (if not all) types of
hardware could benefit from it. Anything else is just that...a custom
add-on to suit the needs of a specific individual. It only makes for more
confusion for those who may not share your expertise. Of course, this
precludes the idea of making the journaled filesystem a blade, same with
the second hard drive backup idea.
Or...as a community we could refine the hardware requirements that e-smith
has put forth so that we all could build boxes with off the shelf hardware
that provide the protection we need without having to modify the software.
What tapes drives have been reported to work, RAID cards, etc, and then a
list of a low, mid and high end box with part numbers so we can be sure
they work. The various Linux databases out there are good, but none to my
knowledge reflect e-smith specifically. A good example is the TrippLite UPS
I bought. Open source software for Linux included. Just don't call the help
desk. Oh, and if you don't have X Windows on the machine, forget about
configuring it. That is why the contrib HOWTOs are so useful. they show us
workarounds. What we need is the hardware equivalent.
Anyway, just my opinions. I could be wrong.
Bob
"Jaime Nebrera" <[EMAIL PROTECTED]> on 09/07/2001 12:01:26 PM
To: <[EMAIL PROTECTED]>
cc:
Subject: [e-smith-devinfo] What do you think?
Hi all,
I just want to post some "ideas" in this list to see what other fellowers
think about them. We plan to invest some time developing for e-smith and
what to get this right.
First a little feedback. I'm a Industrial Engineer and my speciallity is
quality management. My friend (Juan Jesus) is also an Industrial Engineer
and his speciality is electrical systems.Our first contact with e-smith was
a year ago while searching for a solution for a small office we worked at.
They had 4 Windows client and wanted shared acces to the internet, files,
etc, etc. At that moment we didn't know about e-smith so we made a custom
SUSE install for them. It has worked wonderfully for them in this time but
has a big problem, you need to know linux to administer it (webmin is too
difficult for some people :(
So we found e-smith. I believe it's a great product, probably the best
distro arround for such a task, but in this time we have found some minor
"weaknesses". Probably we have found them because we have seen many other
distros, and sure the owner doesn't know they even exist but, well we like
what we made with that SUSE and whant to try to get more or less the same
with e-smith.
First, some times e-smith might be too simple to install. Yes, I agree
its
perfect when you don't know anything about computers and linux but, what
happens when you know what you are doing? This simplicity makes your work a
little harder. One thing thats easy to solve its e-smith way of just using
1
hd and partition it in a hard way. In our custom system we have one
partition just for data and a second (smaller) hd just for backup. A script
runs everynight and makes a complete compressed backup of the data. Yes
sure, you can do this with e-smith downloading to a separate computer or
using a tape unit, but you know most of the times this kind of busnisses
don't have a single procedure for this (for example in our case, the owner
has just made 2 backups in a year from the secondary hd to cdrom when he
was
supposed to do it monthly). Messing with tapes or just remembering to do a
backup sometimes is too much for a unknownfull person. The backup change is
quite easy to implement in e-smith (just change fstab and make a cron
entry)
but not so the data partition (without using a secod hd I mean).
Then comes the filesystem. I have read the faq and know what you think
about other filsystems than ext2. I might agree with you they might be not
stable enough (even when SUSE uses Reiser for a long time without problems
(they were caused by VIA chipset instead)) but as a SUSE user for a long
time (my destop computer has SUSE 7.0) I just love it and have not had a
single problem with it. We would like to try to add some kind of complete
journalling filesystem support for e-smith (not as a separate hd as i have
said). The main reason, well a guy that is scared of a tape unit is also
scared of a UPS system :)))) Seriously, not everybody has a UPS unit and I
don't like the idea of a system that is uncapable of starting up (a broken
system that starts up and connects to the internet might be remotely
solved,
but if it cant start there is no remote chance). Why not givin the rebuild
orders manually? Try to say to a Windows 98 "expert" he should type
something like fdisk -tescsde /dev/hda6 bla bla bla (just kidding) just to
start the computer. He might accept a 2 hour delay to start but sure as
hell
I don't what to get a call saying "My system wont boot up" and have to
travel 500 km to get it solved. I'm talking about robustness. Maybe a
solution could be a separate partition just to make an emergency startup
but, who knows.
We thought about using Reiser because is the one used longer in Linux and
has support for a 2.2.18 kernel (the one used in e-smith). I thought ext3
was too risky yet as is JFS (a suspicius jump form 0.2 to 1.0). Maybe XFS
but I hought it only worked on the 2.4 series kernel. If anybody in this
list has started a similar project, please tell us so we can work together.
If somebody whats to work with us, raise your hand :)
Next comes our monitor package. We are trying to achieve an e-smith addon
(or should I call it blade now :) that lets you "see" the state of the
system. As a quality management specialist I know of the importance of
seeing a tendence to discover when something is wrong. As someone posted in
a previous topic, an alert system would be a great add on for the monitor
system, but this will have to wait jus a bit. Our idea is: let the system
call you when something goes wrong and then use the visual display and
Charlie's log viewer to solve it (thanks Charlie). Why not using Mitel's
service link? Well, with those prices I find it really hard to sell a
single
unit to somebody that just says a UPS unit is expensive and not needed.
Here
in Spain you will find it really hard to sell it (we are the second country
in the world in software piracy :(((( not to mention the lack of spanish
support (at least an internalization support so we could just translate it
in a elegant way, not just editing the functions). BTW, does e-smith have a
lm_sensors aware kernel?
So our future directions would be: finish e-smith monitor (make it stable
and bug free), add an alert system and try to add a journalling filesystem
support. If somebody has anyexperience or wants to help, say it.
Regards.
--
Please report bugs to [EMAIL PROTECTED]
Please mail [EMAIL PROTECTED] (only) to discuss security issues
Support for registered customers and partners to [EMAIL PROTECTED]
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
Archives by mail and
http://www.mail-archive.com/devinfo%40lists.e-smith.org
--
Please report bugs to [EMAIL PROTECTED]
Please mail [EMAIL PROTECTED] (only) to discuss security issues
Support for registered customers and partners to [EMAIL PROTECTED]
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
Archives by mail and http://www.mail-archive.com/devinfo%40lists.e-smith.org
