[pve-devel] faster live migration : hpn-ssh build tutorial for wheezy

Stefan Priebe - Profihost AG s.priebe at profihost.ag
Fri Jul 26 10:49:16 CEST 2013

Am 26.07.2013 10:37, schrieb Dietmar Maurer:
>> -----Original Message-----
>> From: Stefan Priebe - Profihost AG [mailto:s.priebe at profihost.ag]
>> Sent: Freitag, 26. Juli 2013 10:36
>> To: Dietmar Maurer
>> Cc: pve-devel at pve.proxmox.com; spriebe
>> Subject: Re: [pve-devel] faster live migration : hpn-ssh build tutorial for
>> wheezy
>> Am 26.07.2013 10:23, schrieb Dietmar Maurer:
>>>>> But consider the following (this can also happen with current code):
>>>>> - kvm crash
>>>>> - another migration starts shortly after using the same port now
>>>>> - the other side connect now to the wrong VM
>>>> But this can happen with the tunnel too? What do i miss?
>>> yes (as I wrote above)
>>> Anyways, we want to prevent such things?
>> I started to rethink about it and still don't get it.
>> If the source VM crashes and the new "target" vm is still running
> The other way. The target crashes (an another target is started using the same port afterwards).

ah OK. But when the target crashes and a new migration is started - the
new migration will still get a NEW port due to the expiretime logic in
PVE::Tools - the old port isn't exired.


More information about the pve-devel mailing list