Topic: Software RAID follow-up

[Hans Mulder]

After quite a bit of effort, I finally got SW RAID-1 running using e-smith. I have two identical disks that each have identical boot, swap, root, and home partitions. The home partitions run in RAID-1 configuration.

When I subsequently tried to get the root partitions to run in RAID also, I ran into a problem which is most likely caused by the fact that the RAID support is dynamically loaded in the kernel and not compiled in. The problem exhibits itself as a kernel panic (not able to mount /) when booting from the (degraded) RAID root device. Mounting the RAID device later after booting from the original root device works fine.

My question is whether there is an e-smith linux kernel with the RAID support compiled in? Or, if I really want this (can also manually copy root occasionally) do I need to install red-hat instead.

Hans

[Damien Curtain]

Make an initial ramdisk, ie.
mkinitrd /boot/initrd-.img

and add it to lilo with a line:
initrd=/boot/initrd-.img

man initrd will tell you more.

--
 Damien

[Richard Ford]

Red Hat?

It is getitng taht way isn't it??

We are an e-smith authorised partner now.  But I am finding too many restrictions too quickly.  IMHO I beleive essential system stuff should ahve been focused on first - before stuff like Ibays and samba.

Propper RAID and backup support.  Ability to turn services on and off.  We do not need atalk for example.

Setting up all the services that e-smith does manually - is not hard.  If one coudl master DOS and DR DOS back in 1987-94.  And hence windows and NT.  Linux with the same access to books and FAQ's/Howto's is not =that= hard anyhow.  People are just too lazy to learn.

This is where e-smith helps us - repition.  We used to just make a good RH setup and then ghost it to new machines.  E-smith ahs saved a lot of stuffing around....... but it is tredding a fine tightrope.  IMHO too much effort has been placed on ease of use and not enough on making it a robust solution.  Robust in the sense of "server" attributes.  I would forgoe the pretty interface for more functionality in the blink of an eye.

We are trying to elicit information from 3ware - we are about to start a few major roll outs.  And unless we can solve the RAID and security/backup problems, we may have to only use e-smith as a gateway and mail and not the PDC.

Also without RAID and easy backup - we will ahve to forgoe using IMAP on the email box as IMAP was intended as our plan for client email backup.  Combiend with a homogenous desktop environemnt and norton ghost - we have a very low % of project costs associated with maintenance.  A far sight lower than the 69% average they used to thrust down our throats at uni.

As I said - it is on the edge..... we hope we can overcome these problems and not have to mix our server environemnt between NT and e-smith or more likely, e-smith and RH.

Cheers,
Richard.

[Joseph Morrison]

Hello Richard,

Thanks for your comments. We're pleased to have you as an authorized partner!

Richard Ford wrote:

> IMHO I beleive essential system
> stuff should ahve been focused on first - before stuff like
> Ibays and samba.

We've been very successful by listening to our business partners and customers. We implemented Samba from day one because Windows compatibility was in demand right from the start. Then users started asking for intranets, extranets, ftp sites, and so on - so we invented i-bays to meet their requirements.

Earlier this year , as we started increasing our profile in the marketplace, we started getting requests for - guess what? - RAID, backup, and the ability to turn services on and off. We've listened to those requests and have already been developing those features, along with many others, over the last few months. We are not quite ready to announce the next version of e-smith, but stay tuned. You won't be disappointed.

> This is where e-smith helps us - repition.  We used to just
> make a good RH setup and then ghost it to new machines.
> E-smith ahs saved a lot of stuffing around.......

Thank you! That is very much what we're trying to accomplish. To make it more simple and efficient to deploy servers.

> but it is
> tredding a fine tightrope.  IMHO too much effort has been
> placed on ease of use and not enough on making it a robust
> solution.  Robust in the sense of "server"
> attributes.

For us, simplicity is our highest priority, since we feel that there are too many over-complicated server solutions in the marketplace. Our goal is to enable simple, efficient deployment of crashproof servers, which means that we have to be very selective when adding new features. We are listening closely to our users (especially our authorized business partners) to decide which new features to implement, and can't help feeling pleased that everything you asked for is already planned for the next version of e-smith.

As an authorized channel partner, you have a contact person at e-smith dedicated to making sure that you have what you need to deploy network solutions efficiently. Please keep in touch with us, so that we can make sure our software continues to meet your needs!

Best regards,
- Joseph Morrison, CEO

[Richard]

G'Day Joseph,

The new version of e-smith certainly sounds like good news.  e-smith is a great product, yet I have refrained from doing any frankenstein work on the system.  By this I mean manaully tweaking and altering/adding/removing functionality.

This is where the cause of my concerns start.  One wants to use and take advantage of the extra features one needs.  But to do so will break what I understand as the whole "e-smith philosophy".  The more you stuff with it the more likely it will break.  The template system, ease of use, central point of configuration, etc, etc.

And I guess this is the crux of the matter.  One can embrace and use and enjoy the extreme ease of use and speedy deployment.  But the cost to that is the ability to go and manually alter things drastically.  I guess until features are officially introduced - or e-smith packages made available, one will have to decide what is more important?  That is my 'tightrope" analogy - to whcih side do you fall when you need to.

As I said - I am reluctant to fall on the side of excessive manual alteration.  As that is counter productive to what e-smith is trying to achieve.

A friend of mine who is responsible for over 1,000 desktops at the local university is evaluating e-smith now.  He will be pleased to hear that RAID and backup support is coming.  That was his main concern also in e-smith as an option.

Thanks for the response,

Cheers,
Richard.

[Charlie Brady]

Richard wrote:

> And I guess this is the crux of the matter.  One can embrace
> and use and enjoy the extreme ease of use and speedy
> deployment.  But the cost to that is the ability to go and
> manually alter things drastically.  I guess until features are
> officially introduced - or e-smith packages made available, one
> will have to decide what is more important?

I'll have to disagree with you there Richard. The architecture of the e-smith software is designed to make it easily customised, and easily extended with new features. It just takes a little bit of discipline to make the changes in a reproducible and controlled fashion, either through the creation of custom templates, or the creation of add-on packages (and we provide both documentation and a skeleton package to assist the creation of add-on packages).

Software RAID is an exception, as it needs to be setup at system creation time.

If you'd like to discuss custom changes and how to go about making them, the developer mailing list exists for that purpose. Feel free to join in there.

Regards

Charlie

[Richard]

G'Day Charlie,

I certainly will join the list.

I still feel torn between the two ideals?  I do not want to compromise the e-smith system.  I was not entirely referring to the process of adding features, more over the overarching logic of it.  I remember reading a while back in the forums.  Of someone suggesting that we shouldn't be going around throwing in extra features willy nilly, or we won't go around adding them willy nilly.  As this does defeat the ideal of simplicity?

I guess I should ask then, as developers of e-smith.  What was the main use that was intended for the template system.  To clarify did you have any preconceptions as to what people should or should not be adding.  Or was it simply to provide a flexible and systematic architecture for any number of possible future featurres in addition to your own feature development?  

Am I wrong to use it only as a complete off the shelf solution.  Or should I have less reservations and be attempting add new features?


Cheers,
Richard.