On 25.09.20 10:35, Alexandre DERUMIER wrote:
>>> but how do you detect pending changes now?
> 
> Well, the feature was mainly to detect pending change after reload. 
> if a reload don't have applied correctly on a node, or if a node was down. 
> 
> I don't known if we want to display to user "pending config" changes, not yet 
> applied ?

I'd like to have that.

> 
> Befor this commit, It's displaying warning after any config change, 
> and it's difficult to known if a problem occur after the reload. 

On 25.09.20 10:39, Alexandre DERUMIER wrote:
> also,
> 
> for example, when you add a new vnet in a zone,
> 
> it was displaying a warning all vnets/zones for pending changes.
> 
> as I don't have enough granularity currently (a global version info in 
> /etc/network/interfaces.d/sdn, or we should have some kind of versioning info 
> by vnet in /etc/network/interfaces.d/sdn)
> 
> 

Having two versions, the enacted and a pending, could be enough

* if both are the same all is applied
* if pending is newer we can show it, but new changes should not further
  increase the version, they are seen as part of the current pending stuff.
* if pending is older, bug but don't care?

So on each change we bump $pending to $enacted + 1 (*not* $pending++) after we
wrote the changes out. We could make /etc/pve/sdn/.version more structured, 
either
json map or something like:
enacted=3
pending=4

(json could be more flexible)

An apply sets $enacted to $pending once finished (without errors).

This would be simple, not much to track but still give the admin info if 
anything
is pending. What do you think?



_______________________________________________
pve-devel mailing list
[email protected]
https://lists.proxmox.com/cgi-bin/mailman/listinfo/pve-devel

Reply via email to