Here are some improvements ideas

[galphanet]

Hello,

I'm using proxmox VE for some time now and I really like this projet.

So here are some ideas I have for the bare-metal system:

- Add snmp support on the host (and maybe some pretty graphs)
- Add IPv6 support for both (host and VMs)
- Stabilize the network because for now loadbalancing and failover isn't working well
- Remplace Apache with lighttpd or other lighty webserver and the ability to change the server's certificate (but we can do it with ssh)
- Add a warning on the web interface when an update is available (or use apt-cron)
- Add the partition scheme choice when installing the server and/or resizing on-the-fly and reduce the size of the root partition (5-10 Go are sufficient I think)
- Add DRBD support with active-active failover
I mean: having two server with VMs in load balancing, if we lost one server the other will take over the VMs that were on the first.
- Add restore-point solution and duplicate VMs
I mean: for example when I update my VMs or try something, it would be very good to make a snapshot before doing anything in case of problems and/or duplicate it and make tests on the other VM
- Add the choice where the VM would be stored (on which disk, which partition, which server, etc..)
- Add an option to add/remove users and give them rights to manage some VMs
- Add an API solution for managing VMs and clusters from others (homemade) softwares

I know it's very hard to implement some of my ideas but I'm ready to work and help to the development.

Again, many thanks for your work and I hope the V.2 will be the best opensource linux virtualization tool !

 

[benphelps]

I agree with replacing Apache with a more light weight server. I think the only problem will be with the htaccess.

 

[dietmar]

I agree with replacing Apache with a more light weight server. I think the only problem will be with the htaccess.

Really? Do you have some experience in running embperl with lighttpd?

 

[benphelps]

They both support fcgi, didn't think much more about it past that.

 

[dcalvache]

I big improvement would be (in the case that next proxmox supports external storage) that proxmox offers a checkbox and a botton to start automatically and manual the virtual machine on the best host available.
I mean a little script that scan the resources on each host and select automatically the host with the less loads.

Off course it will be great, that if a host with "n" virtual machines will go down, the virtual machines go up automatically in another/another hosts.

 

[ddarko]

1. "Stabilize the network because for now loadbalancing and failover isn't working well"
2. Remplace Apache with Nginx
3. Much more statistics: the number of processes, the use of network ...
4. History of statistics, notifications

 

[dietmar]

Remplace Apache with Nginx

Why?

 

[ddarko]

Why?

1. is very lightweight and easy to configure
2. supports perl
3. is much more efficient than Apache
4. i use it as a web proxy and loadbalancer for VE containers (Apache can't forward, cache, and LB site traffic from the port 80 HN into several containers VE)

one lightweight program instead of a proxy to handle VE containers (wich in my case is require) and PVE (Apache part).
Google has a lot of tests: "Apache vs. nginx".

 

[dietmar]

You claim that is its 'very lightweight' and 'much more efficient'.

I simply do not believe that perl is more lightweight or more effieient when running with nginx - why do you think it is?

Also, do you have some numbers for above claims?

- Dietmar

 

[foxxx]

being able to select a few nodes in a cluster - to turn them into a cloud would be nice too

 

[xRay]

You claim that is its 'very lightweight' and 'much more efficient'.

I simply do not believe that perl is more lightweight or more effieient when running with nginx - why do you think it is?

Also, do you have some numbers for above claims?

- Dietmar

I'l strongly second that, I administrate RT3 which is also Perl based and using NGINX in combination with fastcgi has brought down memory usage of the webserver + perl (apache embedded, nginx + fastcgi) from 250MB down to around 120MB.

nginx is also faster and the system feels, under load, more smothly.

Personally I replaced every Apache occurence where possible (except for webhosting) with nginx (Nagios, RT3, Coldsanic, etc.) because in an environtment like our virtualization system memory counts and you will see benefits if you use lightwight stuff for this sort of things where apache is clearly an overkill.

I am an advocate of using the right tool for the job and as i am new to proxmox (installed yesterday my first testmachine) it was one of the first things that popped into my mind (replace it with nginx).

best regards, and keep up that good work,

Raimund Sacherer

p.s. to your question regarding why it should be lightwighter, one thing is that apache itself alone takes a lot of memory, like 30MB or more unconfigured, whereas nginx uses around 2-3MB unconfigured.
The problem with embedded Perl is because apache loads per definition x servers (like, 3 or 5, whichever is configured) and every instance consumes considerable memory due to the way how perl is integrated into apache.

it's great (apache+perl) for traffic-websites, but for something like a management interface it's overkill.

 

[dietmar]

I'l strongly second that, I administrate RT3 which is also Perl based and using NGINX in combination with fastcgi has brought down memory usage of the webserver + perl (apache embedded, nginx + fastcgi) from 250MB down to around 120MB.

That is VmSize (not RSS). So how much memory do you save really (in terms of RSS)?

- Dietmar

 

[dietmar]

I'l strongly second that, I administrate RT3 which is also Perl based and using NGINX in combination with fastcgi has brought down memory usage of the webserver + perl (apache embedded, nginx + fastcgi) from 250MB down to around 120MB.

Also, those numbers reminds me - here, one perl process need about 120MB, so you could explain above difference that apache starts 2 processes by default (when you use mod_perl), whereas nginx X only starts one as default (because it uses fastcgi)?

 

[xRay]

Hi,

i am talking about Rss memory, not about virtual nor shared!

and i am counting the apache versus the nginx + fastcgi (mason_handler.fcgi) as a matter of fact

best

 

[dietmar]

OK, sounds promising - i will put that on my task list, I will do some test myself ;-)

 

[xRay]

Also, those numbers reminds me - here, one perl process need about 120MB, so you could explain above difference that apache starts 2 processes by default (when you use mod_perl), whereas nginx X only starts one as default (because it uses fastcgi)?

yes, exactly, well, more or less, apache has the overhead due it's nature of embedding the perl engine, so each apache server (i am not sure, can you run embedded perl in a threaded apache server? ...) has to have it's own perl engine, which holds each the compiled scripts, etc. This is very fine and usefull for a real production traffic website where every new fork for a new server hurts if there are no spare ones, etc.)

on the other side, nginx talks to a defined fast-cgi process, so, (i guess nginx is threaded) each of nginx processes just connect to the fast-cgi process and this process is responsible for shareing the connection-datas ...

 

[xRay]

OK, sounds promising - i will put that on my task list, I will do some test myself ;-)

What also would be really great is an option for each VM to be able to hook into the backup process, like:

OnBeforeBackup execute script
OnBeforeSnapshot execute script
OnAfterSnapshot execute script
OnAfterBackup execute script

personally I only would need before and after snapshot, but I think the others (and eventually some more, which I am not thinking of right now) would be nice to have.

why the onBefore and OnSnapshot scripts?
-> e.g. for an mysql host, i can tell before the taking the snapshot to lock the databases, then it makes the snapshot, and as soon as the snapshot is created, I can tell mysql to unlock the databases and ensure with an absolutely minimun on "service-downtime" that I have really a consistent database (especially since standard mysql does not have transactions).

in this szenario the databases get locked for exactly the time needed to create the snapshot, which normally is a whizz...

best regards

 

[svenr]

Hi there,

i´m currently testing proxmox within citrix-xenserver and thinking about replacing xen with proxmox. Because running in VM i only tested the openvz features. I haven´t found a request tracker or wishlist so here are some idea´s ;-)

Networking:
- names for networks (vlans/bridges)
- more then 1 network per vm
- editing / adding / deleting networks of existing vm´s
- using different bridge for systemmanagment an vm´s

Backup:
- direct backup and restore of vm (backup now)
- list of available backups
- limit number of backups to keep
- pre/postscripts
- maybe differential backups

Config:
- ntp-server
- mailserver / senderdomain

Cluster
- versioncheck between clusternodes
- syncing networks between clusternodes
- syncing shared storage between clusternodes (iso/backup etc)
- syncing backups (making shure the vm is backuped after moved to another server)
- maybe logical clustes with identical network/storage/backup-settings
- maintenance-mode for a node (all vm´s are moved to other servers to take down the node for maintenance and afterwards the vm´s are moved back)

Frontend:
- direct link to vnc-console at VM-list
- list iso´s from remote-storage under ISO-Images

also a tracking of configchanges and comment-fields would bee nice.


Sven

 

[tom]

Hi there,

i´m currently testing proxmox within citrix-xenserver and thinking about replacing xen with proxmox. Because running in VM i only tested the openvz features. I haven´t found a request tracker or wishlist so here are some idea´s ;-)

Networking:
- names for networks (vlans/bridges)
- more then 1 network per vm
- editing / adding / deleting networks of existing vm´s
- using different bridge for systemmanagment an vm´s

Backup:
- direct backup and restore of vm (backup now)
- list of available backups
- limit number of backups to keep
- pre/postscripts
- maybe differential backups

Config:
- ntp-server
- mailserver / senderdomain

Cluster
- versioncheck between clusternodes
- syncing networks between clusternodes
- syncing shared storage between clusternodes (iso/backup etc)
- syncing backups (making shure the vm is backuped after moved to another server)
- maybe logical clustes with identical network/storage/backup-settings
- maintenance-mode for a node (all vm´s are moved to other servers to take down the node for maintenance and afterwards the vm´s are moved back)

Frontend:
- direct link to vnc-console at VM-list
- list iso´s from remote-storage under ISO-Images

also a tracking of configchanges and comment-fields would bee nice.


Sven

thanks for sharing this. (just to mention, quite some parts are already done and released in the current stable.)

 

[svenr]

thanks for sharing this. (just to mention, quite some parts are already done and released in the current stable.)

I used 1.4 with latest updates, but only VZ Part (no VT). With network-names i think about something like an alias. like 'dmz2 costumer xxx' instead 'vmbr142'.


btw great tool, I'm looking forward to the next version