Koozali.org formerly Contribs.org

High Availabilty (Fail-over, Fail-back) with Raid-1 between

guestHH

High Availabilty (Fail-over, Fail-back) with Raid-1 between
« on: February 10, 2002, 12:35:45 AM »
Hi all,

I'm busy with creating High Availablity for 2 SME servers bases upon Heartbeat (http://www.linux-ha.org) and DRBD (http://www.complang.tuwien.ac.at/reisner/drbd/)
. Let me explain;

- 2 SME servers represented to the user as only 1 server (with exact same data on both of them)
- Automatic Fail-over, Fail-back between the 2 of them (with current data)
- real- time mirroring (RAID-1) of user data (incl. open files e.g. databses) from server 1 to server 2.

I'm getting the point where it's going to be usefull in practice, so I'm would like to know if there are any of you really interested in a draft How-To.

Note this, I'm not the same caliber as other guys like e.g. Dan B., Darrell, Des, Charlie, Jamie or Filippo and MANY others, but since I had the need for this HA 'feature' I just gave it a try......

I would like to know if you would be interested in a How-To that will get you started so that I know if I have to translate my personal (unreadable) notes in a more clear paper AND i would also like to know how 'big' the demand for such a feature would be so maybe others are interested to improve the initial how-to for there are many issues to be resolved to get to a 'out of the box' solution. (and maybe Mitel likes the idea  too 'cause of the 'high' demand...:-)...)

So, please reply with 'yes' or 'No' to this thread so we ca get a better picture of the 'need' for this. Maybe you would like to add the exact situation you would like to use it for or any other comment, It would be appreciated.

If you think 'what's this? what a crap..' that's ok by me, please ignore this posting.

Regards,
guestHH

Kelvin

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #1 on: February 10, 2002, 12:37:30 AM »
Hi HF,

YES ! YES ! YES !

Please post a How-To for us to try !

Cheers,

Kelvin

Peter

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #2 on: February 10, 2002, 12:49:44 AM »
Yes and when your ready we will put two Test servers together and test it for you

Tom Keiser

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #3 on: February 10, 2002, 12:57:16 AM »
RequestedDeletion:

    I have one customer who is very interested in this concept. I am building them a *second* server, both servers identical with raid 1 controllers, and duplicate removeable data drives and duplicate boot drives (total of 4 removeable drives in all). Only server 1 or server 2 would be operational at a time, and if the online server failed, the admin would shut it down, transfer the removeable data drive carriers plus one of the boot drive carriers to the other server and boot it up. The extra boot drive would be an exact copy of the main boot drive for use in case of data corruption or rootkitting, etc., and would not normally be "live".

    Your HA (assuming it is reliable) would be an improvement, in that the changeover would be more or less instant. It would
involve more hard drives than my solution, but its elegance and automatic operation should offset that.

Regards,

Tom

steve

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #4 on: February 10, 2002, 01:04:15 AM »
I would like to give it a try. Do machines have to be same spec ?

I have 750 Duron 80 GB & a P75 1 GB. The duron is email and a few page web server for home + 3,200 MP3's. Whilst i cant mirror the MP3's i should have back email as i dont have any ISP service i host my own.

Steve.

guestHH

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #5 on: February 10, 2002, 01:18:31 AM »
Hi,

FYI and further reference:
Server 1 is named Goliath (active, primary server ip 192.168.1.201) Server 2 is named David (Stand by, not actibe but powered on server, ip 192.168.1.202)
Both servers can NOT be seen on the network via network neighborhood (but pingable)
The 'cluster' of both servers is called 'The Rock' and is present on the network and is viseable to the users as 'The Rock' and has 'virtual' ip 192.168.1.210 (virueal ip provided by the HA system)

Kelvin: I heard you ;-)

Peter: Thanks in advance I would be interested in nay feedback

Tom:
Your solution is clever. Just to give you more info at this moment:

I use 1 small system (e-smith system(drive and  lare data drive (both IDE) on each server (installed and mounted AFTER installation of e-smith.
Current status is that user data is real-time mirrored to server 2 (stand by) over a 2nd NIC on a 'private' net 10.0.0.0. Data saved to The Rock will be Real-Time mirrored to David (stand by) with a rate of 1.8-1.9 Mb/s, 100mb NIC)

And yes realiability is very important. That's why i posted the initial message and invitaion to paticipate and find out to what extend.

Regards,
guestHH

guestHH

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #6 on: February 10, 2002, 01:26:16 AM »
Hi Steve,
No the servers don't have to be of the same specs. The ONLY restriction is that the amount of mirrored (LAN RAID-1) data fits on both user-data drives on both drives. (server A 10Gb server B 5 Gb won't work)  

That's it. BUT it helps the performance that you use 7200 rpms drives, 100Mb nics, Processor speed as high as possible (but don't need to be the same on the 2 servers) and connect the servers to separate power outlets.

The 2 servers can be max. 80 meters apart (in case of fire etc.... whatever)


For you case, a friend of mine is now using a 2nd (give away from another friend) server that is mirroring his MP3's ... :-)

Regards,
guestHH

Jeff C

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #7 on: February 10, 2002, 02:21:48 AM »
HF,

Got your links on the older thread and will now catch up on this one.  Thanks!

As you are aware, I'm very interested in a howto on this subject and I can dedicate some hardware and time to testing.  I'd be happy to test your howto and help in any way that I can.

Cheers,

-jeff

guestHH

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #8 on: February 10, 2002, 02:25:53 AM »
Hi Jeff,

Thanks in advance.

Since I live in th southern part of The Netherlands, we celebrate Carnaval over here very intense !!! The how-to will probebly will arrive somewhere next week.

(if i get over my headache.........)     :-)

Regards and thanks,
guestHH

jeff C

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #9 on: February 10, 2002, 05:48:33 PM »
Hoist one for me!

-jeff

René

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #10 on: February 11, 2002, 07:00:46 PM »
Hi RequestedDeletion

Yes, please post a how-to to setup the sme to a high availabilty server. That's realy what we need.

Thx

Kind regards

René

Simon

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #11 on: February 12, 2002, 02:13:49 AM »
I would also be VERY pleased to read such a thing.

Regards

guestHH

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #12 on: February 12, 2002, 02:40:30 AM »
You're on the list Simon :-)

Regards,
guestHH

John Lewis

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #13 on: February 12, 2002, 03:09:03 AM »
I am very interested in this as well.

High availability is a very popular topic these days, and having worked in datacenters, I see it in use all the time.  Making it available for the SME server would greatly enhance its ability to get into businesses

guestHH

Re: High Availabilty (Fail-over, Fail-back) with Raid-1 betw
« Reply #14 on: February 12, 2002, 03:49:25 AM »
Hi John,

Added your address to the list... ;-)

I'm finishing the How-To, it will arrive to the end of this week so all of you can have a go at this during the weekend :-) so get your hardware ready:

- 2 servers with 2 harddisks with 2 NICS, 1 null-modem cable

- 1 small 'system' harddisk (10 Gb?) and 1 large data disk (40+Gb)
- Make sure you have internet access to download RPM's

*** DRAFT HOW-TO, FOR TESTING PURPOSES ONLY ***

Regards,
guestHH