Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Kevin Lemonnier]

> But if using network backed storage you could still have lots of free
> CPU and RAM but still see VM's running in slow motion due to network
> congestion. 
> 

Yes, that's true, and if you are doing HA you have to be using a networking
backed storage. That's a good point !
It never happened to us yet, we have plenty of bandwith left, so I hadn't
considered that.

-- 
Kevin Lemonnier
PGP Fingerprint : 89A5 2283 04A0 E6E9 0111


signature.asc
Description: Digital signature
___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Michael Rasmussen]

On Tue, 18 Oct 2016 12:41:25 +0200
Kevin Lemonnier  wrote:

> 
> I had not considered network bandwith, that would indeed be interesting .. 
> Seems a bit less
> easy to do though, cpu and ram are the big things I think, at least ar first.
> 
But if using network backed storage you could still have lots of free
CPU and RAM but still see VM's running in slow motion due to network
congestion. 

-- 
Hilsen/Regards
Michael Rasmussen

Get my public GnuPG keys:
michael  rasmussen  cc
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xD3C9A00E
mir  datanom  net
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE501F51C
mir  miras  org
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE3E80917
--
/usr/games/fortune -es says:
This is your fortune.


pgpv6knkjo9kk.pgp
Description: OpenPGP digital signature
___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Kevin Lemonnier]

> 
> Would be great. Also note that automatic load balance should only be possible 
> for
> VM's fulfilling requirements for running as HA. This means HA pools can
> be taken into consideration when moving a VM.
> 

I imagine, yes. We don't use this at all, we don't mix HA and non HA VMs on the 
same clusters,
so I can't really talk for that.

I had not considered network bandwith, that would indeed be interesting .. 
Seems a bit less
easy to do though, cpu and ram are the big things I think, at least ar first.

-- 
Kevin Lemonnier
PGP Fingerprint : 89A5 2283 04A0 E6E9 0111


signature.asc
Description: Digital signature
___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Michael Rasmussen]

On Tue, 18 Oct 2016 12:28:12 +0200
Kevin Lemonnier  wrote:

> 
> Yes, something like VMWare's DRS would be great to have. For now I "load 
> balance" by hand when I get a load alert
> on a host, or when a client complains that the VM is slow. I was thinking of 
> automating that through the API,
> I already have an external tool to manage routing the IPs depending on where 
> the VM is running, guess it'd be
> easy enough to query the ram or CPU and try to move VMs, but that'd require 
> some heavy testing that I don't have
> the time for right now.
> It's on my todo though, as long as Proxmox doesn't come up with something 
> similar in the meantime.
> 
Something like this has been discussed in the past. IMHO a load balance
mechanism based on:
1) CPU cores
2) Memory
3) Network bandwidth

Would be great. Also note that automatic load balance should only be possible 
for
VM's fulfilling requirements for running as HA. This means HA pools can
be taken into consideration when moving a VM.


-- 
Hilsen/Regards
Michael Rasmussen

Get my public GnuPG keys:
michael  rasmussen  cc
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xD3C9A00E
mir  datanom  net
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE501F51C
mir  miras  org
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE3E80917
--
/usr/games/fortune -es says:
Mr. Rockford?  You don't know me, but I'd like to hire you.  Could
you call me at...  My name is... uh...  Never mind, forget it!
-- "The Rockford Files"


pgpRF2jPo4l2y.pgp
Description: OpenPGP digital signature
___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Kevin Lemonnier]

> Imho it is not relevant to aggregate ressources of the whole cluster, because 
> there is no mechanism of load balancing.
> So you could have for ex : some nodes almost full of RAM, and other nodes 
> half empty..
> 
> For now i workaround the problem always creating a new VM on the node with 
> fewer number of VM, and add it to HA.
> But it's not ideal, because the number of vm have nothing to do with how much 
> RAM is consumed...

Yes, something like VMWare's DRS would be great to have. For now I "load 
balance" by hand when I get a load alert
on a host, or when a client complains that the VM is slow. I was thinking of 
automating that through the API,
I already have an external tool to manage routing the IPs depending on where 
the VM is running, guess it'd be
easy enough to query the ram or CPU and try to move VMs, but that'd require 
some heavy testing that I don't have
the time for right now.
It's on my todo though, as long as Proxmox doesn't come up with something 
similar in the meantime.

-- 
Kevin Lemonnier
PGP Fingerprint : 89A5 2283 04A0 E6E9 0111


signature.asc
Description: Digital signature
___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Dhaussy Alexandre]

It would be nice to have a unified view with all nodes ressources.
I was thinknig of doing this with an external metric server (graphite/grafana.)

Imho it is not relevant to aggregate ressources of the whole cluster, because 
there is no mechanism of load balancing.
So you could have for ex : some nodes almost full of RAM, and other nodes half 
empty..

For now i workaround the problem always creating a new VM on the node with 
fewer number of VM, and add it to HA.
But it's not ideal, because the number of vm have nothing to do with how much 
RAM is consumed...

Cheers,
Alexandre.

Le 17/10/2016 à 17:09, Kevin Lemonnier a écrit :


do you mean the resources from each node?




No, I mean for the cluster. Would be great to have those values
(cpu, ram, maybe IO ..) for the whole cluster monitored, I think.
Monitoring each node individually is nice, but that doesn't tell
you that you are almost out of ram on the cluster for example,
you could still be adding VMs automatically without realising
the other nodes are full.





___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user


___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Kevin Lemonnier]

> 
> do you mean the resources from each node?
> 

No, I mean for the cluster. Would be great to have those values
(cpu, ram, maybe IO ..) for the whole cluster monitored, I think.
Monitoring each node individually is nice, but that doesn't tell
you that you are almost out of ram on the cluster for example,
you could still be adding VMs automatically without realising
the other nodes are full.

-- 
Kevin Lemonnier
PGP Fingerprint : 89A5 2283 04A0 E6E9 0111


signature.asc
Description: Digital signature
___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Dominik Csapak]

On 10/17/2016 04:07 PM, Kevin Lemonnier wrote:

t to be able to get each values separatly
for monitoring purposes. Right now we monitor each nodes independently,
I would be happy to add a "cluster check" on top of that though !
Would help knowing when it's time to increase the number of nodes.


do you mean the resources from each node?

i only use the /cluster/resources api call
maybe this is what you want?

there you have all resources for each node and vm

___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Dominik Csapak]

On 10/17/2016 04:19 PM, Michael Rasmussen wrote:

On Mon, 17 Oct 2016 16:17:48 +0200
Michael Rasmussen  wrote:


I would say 5 out of 6 cores is in use so 83,33 % CPU usage in the
cluster.


Forgot to mention: For people coming from VmWare this makes sense since
that is how vSphere cluster client displays it.



yes, thinking about it, it does really make more sense, thanks :)


___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Michael Rasmussen]

On Mon, 17 Oct 2016 15:58:35 +0200
Dominik Csapak  wrote:

> so cpu is the node average
Why average on not total? If this is supposed to be a cluster wide
dashboard average gives no mening.

-- 
Hilsen/Regards
Michael Rasmussen

Get my public GnuPG keys:
michael  rasmussen  cc
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xD3C9A00E
mir  datanom  net
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE501F51C
mir  miras  org
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE3E80917
--
/usr/games/fortune -es says:
There must be at least 500,000,000 rats in the United States; of course,
I never heard the story before.


pgpJh13Xl3Khq.pgp
Description: OpenPGP digital signature
___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Dominik Csapak]

On 10/17/2016 04:04 PM, Michael Rasmussen wrote:

On Mon, 17 Oct 2016 15:58:35 +0200
Dominik Csapak  wrote:


so cpu is the node average

Why average on not total? If this is supposed to be a cluster wide
dashboard average gives no mening.



i agree, but with cpu usage it is not so easy

for example:

if you have 1 node with 50% usage and 2 cores
and 1 with 100% usage and 4 cores

what would you display here?


my initial idea would be to display: 75% used of 6 CPU(s) (maybe i 
should change this to cores)


but you are right it would probably be better to display:

83% of 6 cores

or do you mean something completely different?
(keep in mind we only get # of cores and total cpu usage in the api call 
i use)


___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Michael Rasmussen]

On Mon, 17 Oct 2016 16:17:48 +0200
Michael Rasmussen  wrote:

> I would say 5 out of 6 cores is in use so 83,33 % CPU usage in the
> cluster.
> 
Forgot to mention: For people coming from VmWare this makes sense since
that is how vSphere cluster client displays it.

-- 
Hilsen/Regards
Michael Rasmussen

Get my public GnuPG keys:
michael  rasmussen  cc
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xD3C9A00E
mir  datanom  net
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE501F51C
mir  miras  org
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE3E80917
--
/usr/games/fortune -es says:
FORTUNE PROVIDES QUESTIONS FOR THE GREAT ANSWERS: #19
A:  To be or not to be.
Q:  What is the square root of 4b^2?


pgpzxMxkSgWao.pgp
Description: OpenPGP digital signature
___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Michael Rasmussen]

On Mon, 17 Oct 2016 16:14:37 +0200
Dominik Csapak  wrote:

> On 10/17/2016 04:04 PM, Michael Rasmussen wrote:
> > On Mon, 17 Oct 2016 15:58:35 +0200
> > Dominik Csapak  wrote:
> >  
> >> so cpu is the node average  
> > Why average on not total? If this is supposed to be a cluster wide
> > dashboard average gives no mening.
> >  
> 
> i agree, but with cpu usage it is not so easy
> 
> for example:
> 
> if you have 1 node with 50% usage and 2 cores
> and 1 with 100% usage and 4 cores
> 
> what would you display here?
> 
I would say 5 out of 6 cores is in use so 83,33 % CPU usage in the
cluster.

-- 
Hilsen/Regards
Michael Rasmussen

Get my public GnuPG keys:
michael  rasmussen  cc
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xD3C9A00E
mir  datanom  net
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE501F51C
mir  miras  org
http://pgp.mit.edu:11371/pks/lookup?op=get&search=0xE3E80917
--
/usr/games/fortune -es says:
FORTUNE PROVIDES QUESTIONS FOR THE GREAT ANSWERS: #19
A:  To be or not to be.
Q:  What is the square root of 4b^2?


pgpbylTfNc7lA.pgp
Description: OpenPGP digital signature
___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user

Re: [PVE-User] Feedback wanted - Cluster Dashboard

[Kevin Lemonnier]

> 
> https://www.pictshare.net/cb2c08d9ca.png
> 

Looks good, it would be great to be able to get each values separatly
for monitoring purposes. Right now we monitor each nodes independently,
I would be happy to add a "cluster check" on top of that though !
Would help knowing when it's time to increase the number of nodes.

-- 
Kevin Lemonnier
PGP Fingerprint : 89A5 2283 04A0 E6E9 0111


signature.asc
Description: Digital signature
___
pve-user mailing list
pve-user@pve.proxmox.com
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-user