[pve-devel] tasks / ressources

[Alexandre DERUMIER]

>>It does not matter if that is slow - why? 

It was related to my last point for vm status (stop/start) refresh, where it's matter to have fast refresh.

(One other thing not related, it's that an hanging storage can sometime delay/block the others update, 
like node_update, which can give us ta a node red in pve-manager)




>>This would multiply cluster traffic by factor 10?! 

I don't known, is it so big to stream vm status each second ? (just a boolean, stop/start/ 1/0 ) ?


----- Mail original ----- 

De: "Dietmar Maurer" <dietmar at proxmox.com> 
À: "Alexandre DERUMIER" <aderumier at odiso.com> 
Cc: pve-devel at pve.proxmox.com, "Stefan Priebe - Profihost AG" <s.priebe at profihost.ag> 
Envoyé: Lundi 10 Juin 2013 07:06:17 
Objet: RE: [pve-devel] tasks / ressources 

> >>You have up to 16 different cluster nodes, and they update at different 
> times. 
> >> 
> >>But I simply never had problems polling each second. 
> Ok for a big cluster, now if you have a only a single node, it doesn't make 
> sense. 

Maximal time to get new values is 11seconds. This time will increase if you update less often. 

> in pvestatd, 
> 
> my $updatetime = 10; 
> 
> so update status occur each 10seconds. 
> 
> 
> 
> >>pvestatd is a separate processs - so what do you miss? 
> As a storage status can be "slow", do 
> update_qemu_status(),update_storage_status(),update_node_status() in 
> parallel. 
> This can speedup things. 

It does not matter if that is slow - why? 

> > (My main problem is that I can see noticable delays when start/stop vms) 
> >>I doubt this is related to pvestatd? 
> 
> But this is pvestatd which pool QemuServer::vmstatus() to see if the vms are 
> stop/started, right ? 
> So pooling each 10s, you can have delay on status refresh in gui. 

yes. 

> 
> 
> 
> What I would like to have, is: 
> 
> - vm status : pooling each second (pvestatd / pve-manager) 
> - vm,storage,node stats : pooling each X (10?) seconds (pvestatd/ pve- 
> manager) 

This would multiply cluster traffic by factor 10?!