I see that quite a few of you are using H-sphere as a control panel. I've had a look into H-sphere, and it looks quite a good panel.

If you use H-sphere, do you have it installed on one machine, or with different components installed on different machines like Psoft recommend? If so, which components on which machines, and what spec are these machines?

Thanks in advance... just doing a little research.

Matt

I've been using H-Sphere on my server. I think you need at least 256 MB of RAM on the server if you want to install all the components on one server only. My server specs is P3 1 Ghz with 512 MB RAM.

Despite the problems I had on the initial setup, it's been running great and I like it. I heard that the new version 2.1 with much better interface is coming soon. I hope PSoft would get my H-Sphere upgraded to this latest version for free :)

Small note here, if you going to use ony one server, save yourself some hedaches and use 512MB of ram

Anybody its using it in a W2K server?

We have H-Sphere setup ina cluster with a few FreeBSD servers and Win2k servers. I have to say it works great not one real problem as of yet.. Works great on win2k as well.

I'm more interested as using it as a cluster - what kind of cluster do you have setup Rapid? (ie. invidiual web / mail / database servers?)

Thanks
matt

Man i am not sure i want to reply to that past as it was # 666.. LOL

But yes we have 1 DNS / mySQL server, 1 DNS2 CP server, 1 msSQL server, FreeBSD Web Servers, Windows 2000 Web Servers.

We are going to be adding about 20 more servers to the cluster over the next few weeks but we are good with that setup for now. Psoft seems to have a very good idea with this.

The one thing i like is if the mail server goes down SITES stay up. IF a web server goes down the MAIL stay up. Your not putting all your stuff in one place. I really like this idea better the Plesk, CPanel and other CP ideas..


Over all i give H-Sphere a 9. Why a 9 just because it does not support ded. servers all that well yet. Once that is out i say 10 with no problem. But they need to support boht *nix and Windows 2000 Ded. servers.


Let me know if you need any info.. :)

I'm not a host, just a customer. My host uses H-Sphere and I have to say I'm really content with it. I mean we had CPanel before and I thought I can configure and adminstrate many things on my account but that's nothing compared to H-Sphere.
You know, I think it's just great buying one more GB of bandwidth or another 50MB diskspace... maybe add 6-7 more domains... I can do all of it myself.

I just ordered an HSphere install on Thursday, I'm hoping to see something with it soon. I haven't heard back from them yet but I can't wait to see this thing working. A couple questions for those experienced:

1 - Can you have redundant mail servers?
2 - Can you have redundant CP servers?

It seems like setting up redundant mail servers is nothing more than just setting up another mail server to handle store and forward. So far though, I'm totally impressed with the functionality that HSphere offers and it seems like it'll only get better.

I would think you could do Mail not 100% sure on CP but i think you could as it also works off the one db but i can't say for sure as H-SPhere works off domains a lot you may be able to have cp2.domain.com not sure. That is a good idea..

Redundant CP servers are not supported in H-SPhere as of yet, but it is not a big problem, cause if CP server is down. all your other servers are working and you can backup database onto another disk or machine.

Originally posted by voxtreme-matt
I'm more interested as using it as a cluster - what kind of cluster do you have setup Rapid? (ie. invidiual web / mail / database servers?)

Thanks
matt

When you guys mean "clustering", are you talking about redundancy or fail over? Or is it just an H-Spere term menaing a bunch of servers managed by the head of the snake (CP)?

I am curious as in reading the docs, it looks like there is no built in clustering to where I can add two webservers of the same name and if one goes down, the CP sees this and sends traffic to the next. Like simple load balancing.

Thanks.

HSphere uses clustering to distribute work load, not necessarily provide failover or fault tolerant services. By definition, clustering is just a means to promote and support scalability with redundancy as an added bonus. It'll be real cool if PSoft implements such technology on top of what they already provide.

One thing is for sure, you can scale very well with HSphere.

Yea it would be very nice if they made the web servers redundant.

I think this would be very cool. But if they did this I would hope they would give options to as what plans would get setup on a redundant server. This way you could still have like your low end packages on a raid5 system but up sale high end backed with like a 3 Web Server Cluster, Mail Cluster. This would be very nice option to H-Sphere.


Now you could do all this by hand if you wanted to for web servers. It would not be an auto thing but would not be hard at all to do..

Not sure if this would work 100%. But have a FreeBSD server setup and make a copy of the hard drive after the install of H-Sphere. Bring it on-line with a new IP address. Then every 12 or 24 hours make a copy of any to do with users on to the new server.

Now if Server 1 goes down a little script that can edit the IP very fast could bring server 2 on-line with the new IP's.

I am sure there is a better way of doing just never needed to yet with a control panel in mind.. :)

We actually designed it so it would be posible to do something like that using NFS mounted drive.
This is the main reason why we have /hsphere/local & /hsphere/shared directories.

Here is the design:
You have as many servers as you want, mail, dns, web -- all of them working from NFS mounted partitions (both /hsphere/local - user files, and /hsphere/shared - binary filels). You also have 1 or 2 standby serves.

No the trick is: Separate script monitors that servers are up and running. If it detects that there is a problem with one of the servers - it reconfigures one of the stand by servers to mount /hsphere/local & /hsphere/shared directory to replace broken server. After that same script shuts down port on the switch for the proken server and activates brings up IPs on stand alone server (That is also why you can see service IP for each server in H-Sphere's E.Manager).
The whole switchover is under 30 seconds, but it depends on the detection time out - as you don't want to shutdown server if it is overloaded.

Considering that such setup uses single NFS server - it is better be clustered (which is pricy) - to eliminate single point of failure. Now, such configuration requires additional customization, to support disk quota on nfs server, as well as managing switches.

That was our design for for 99.99 reliability.
Not considering NFS server - it is also pretty cheap, considering that you can use single stand by server for 100s of computers.

It should be posible to use stand-by server to provide reliability for some single web (mail/dns) server. Simply by rsync all the data constantly, and having some script to check and fail over.

High availability solution was also designed. That solution would withstand very high traffic, that would not be posible with single server. It has several issues:
1) Requires load balancer
2) Would not work with some software (software that uses shared memory - and as such - cannot be used accross seveal computers). For example built in php sessions would not work, as they track data in memory (correct me if I am wrong).
3) Also requires NFS server (but rsync of data can also be performend)

Such solution would allow any sites on the cluster to handle sites something like "slashdoted" sites, or any other jumps in traffic, without adding lots of extra cost in the hardware.

"NFS"=?

NFS - network file system - it allows several computers accessing the same disk over the network.
You need some kind of network attached storage (NAS) device, that supports NFS.
This can be done by a single linux server runing nfs server, or by something like:

http://www.netapp.com/
http://www.raidzone.com/
http://www.emc.com/

Originally posted by iseletsk

High availability solution was also designed. That solution would withstand very high traffic, that would not be posible with single server. It has several issues:
1) Requires load balancer
2) Would not work with some software (software that uses shared memory - and as such - cannot be used accross seveal computers). For example built in php sessions would not work, as they track data in memory (correct me if I am wrong).
3) Also requires NFS server (but rsync of data can also be performend)

Such solution would allow any sites on the cluster to handle sites something like "slashdoted" sites, or any other jumps in traffic, without adding lots of extra cost in the hardware.

Actually, most load balancing situation will require some amount of custom programming. You mentioned an example of HTTP sessions, it can be solved by storing sessions to a central database server, but the bottleneck will be at the database server. Sure, the database server can be replicated to a few servers, but then the scripts will need some logic. Which db server does it need to contact when doing queries? When doing updates?

The only type of site that won't require custom programming is static sites like download sites, image servers, etc.

Well, clustered databases is even more difficult.
Oracle parallel server (as well as IBM db2, and MS SQL server in clustered setup) cost a fortune. PostgreSQL has rudimentary support for redundancy and MySQL has some support for high-availability - that can be used - but that requires extra programming.

Yet, very often web server can be killed much faster then database server, and in anyway, any such hit can make the rest of the sites on that server pretty much dead - that is why the solution was developed for.