[pve-devel] ha and zfs replication bug

[Alexandre DERUMIER]

>>At the moment this is expected behavior. 

>>When a HA migration is done the source is offline, so we can't migrate the actual state of this replica job. 
Couldn't we change the replication job target, when the config file is move between old and new node ?



>>The replica try to sync a full image what will not work because the image exists on the outer side. 
>>So we need to implement a recover replication state. 
>>It will come soon. 

Maybe add an option in replication job, to force delete of existing target image or rename it as archive (maybe user want to keep datas by security, as replication is async, the old volume has more fresh datas)



>>But I'm not sure about failback, because if the failback will start before the recover of the state is finished (will try every min at 00) the failback can't work. 

for me, the fallback is working, After change the job manually and delete the volume manually.



BTW, I'm thinking to implement replication for ceph rbd next year.(for disaster recovery between 2 ceph clusters on differents sites).
As it's almost same than zfs (snapshot, send-receive),so I think it should work.




----- Mail original -----
De: "Wolfgang Link" <w.link at proxmox.com>
À: "aderumier" <aderumier at odiso.com>, "pve-devel" <pve-devel at pve.proxmox.com>
Envoyé: Jeudi 23 Novembre 2017 07:12:20
Objet: Re: [pve-devel] ha and zfs replication bug

Hi Alexandre, 

> Is it a bug or an expected behaviour ? 
At the moment this is expected behavior. 
When a HA migration is done the source is offline, so we can't migrate the actual state of this replica job. 
The replica try to sync a full image what will not work because the image exists on the outer side. 

So we need to implement a recover replication state. 
It will come soon. 

But I'm not sure about failback, because if the failback will start before the recover of the state is finished (will try every min at 00) the failback can't work.