[pve-devel] [PATCHv2] Add support for multipath-backed direct attached storage

[Alexandre DERUMIER]

>>So question about way to run code on all nodes is still open. Maybe get 
>>node list from PVE and just run what we need over ssh? Or there's some 
>>special API? 

They are not api for that currently.

I think ssh should be better, as maybe sometime api could be unresponsive (hanging daemon for example).

I don't known if it's a possible, but a better way could be to add  multipath and devices manually only 
on nodes where vm is running.

(but maybe this require some udev tricks to not scan devices by default)



----- Mail original -----
De: "Dmitry Petuhov" <mityapetuhov at gmail.com>
À: "pve-devel" <pve-devel at pve.proxmox.com>
Envoyé: Lundi 11 Juillet 2016 12:00:46
Objet: Re: [pve-devel] [PATCHv2] Add support for multipath-backed direct attached storage

Yes, deleting multipath dev itself is just another command ``multipath 
-f $dev'' after (or better before) scsi devices deletion. 

iSCSI is cool, but I don't want to be bound to it because I have also 
SAS and FC storages, so multipath is the only option. 

So question about way to run code on all nodes is still open. Maybe get 
node list from PVE and just run what we need over ssh? Or there's some 
special API? 

11.07.2016 12:12, Alexandre DERUMIER worte: 
> 
> You also need to remove multipath dm devices after deleting. 
> It's really a pain to manage it cluster wide. 
> 
> That's why in past, with my nexenta iscsi san, I was using qemu libiscsi + lacp without multipath. 
> 
> (Also, multipath daemon is really cpu hungry, when you have a lot of luns) 
> 
> 
> BTW, it seem that last version of libiscsi support multipath 
> 
> https://github.com/sahlberg/libiscsi/pull/166 
> 
> But I'm not sure it's already work with qemu. 
> 
> 
> 
> ----- Mail original ----- 
> De: "Dmitry Petuhov" <mityapetuhov at gmail.com> 
> À: "pve-devel" <pve-devel at pve.proxmox.com> 
> Envoyé: Lundi 11 Juillet 2016 10:53:59 
> Objet: Re: [pve-devel] [PATCHv2] Add support for multipath-backed direct attached storage 
> 
> Okay, there's problem: Linux is little ugly when it comes to delete 
> external LUN. 
> So we have to unlink (rescan-scsi-bus.sh -r or echo 1 > 
> /sys/block/sdX/device/delete) all paths to this LUN from all nodes, and 
> only then we may safely delete it. BTW, iSCSI (not direct) storage is 
> also affected (at least if we have installed multipath-tools). 
> 
> Is there easy way to run some code on all cluster nodes, wait it to 
> complete and then continue? 
> 
> _______________________________________________ 
> pve-devel mailing list 
> pve-devel at pve.proxmox.com 
> http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 
> 
> _______________________________________________ 
> pve-devel mailing list 
> pve-devel at pve.proxmox.com 
> http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel 

_______________________________________________ 
pve-devel mailing list 
pve-devel at pve.proxmox.com 
http://pve.proxmox.com/cgi-bin/mailman/listinfo/pve-devel