[pve-devel] gluster patch

Alexandre DERUMIER aderumier at odiso.com
Thu Jan 9 06:43:12 CET 2014


Hi, 

I have done tests with qemu 1.4, and I can reproduce the timeout bug with qcow2 and preallocate metadatas.

qemu-img is crashing ( [qemu-img] <defunct>).

So it must be a bug with current qemu version.


Can you test with qemu 1.7 from pvetest ?

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

De: "FinnTux" <finntux at gmail.com> 
À: "pve-devel" <pve-devel at pve.proxmox.com> 
Envoyé: Mercredi 8 Janvier 2014 21:16:44 
Objet: Re: [pve-devel] gluster patch 


Are 
you using xfs for your bricks? I had those timeouts too but mounting xfs bricks with allocsize=64k helped. 






2014/1/8 Alexandre DERUMIER < aderumier at odiso.com > 


>>Even after a total reboot of the server, the same lock/timeout occurred for both new VM creation and restore. 

I'll do test with qemu 1.4 tomorrow to see if I can reproduce. 

what is your glusterfs setup ? (hardware , raid , replication level) 



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

De: "Ferrara" < aixsyd at gmail.com > 
À: "Alexandre DERUMIER" < aderumier at odiso.com > 
Cc: "pve-devel" < pve-devel at pve.proxmox.com > 
Envoyé: Mercredi 8 Janvier 2014 03:58:05 
Objet: Re: gluster patch 


Both. 


Even after a total reboot of the server, the same lock/timeout occurred for both new VM creation and restore. 


Sam 





On Tue, Jan 7, 2014 at 9:56 PM, Alexandre DERUMIER < aderumier at odiso.com > wrote: 



>>I think I may have found a solution. Its shockingly simple. Theres a line that reads: 
>>push @$cmd, '-o', 'preallocation=metadata' if $fmt eq 'qcow2'; 

Sure, 

but I would like to get it working with preallocation=metadatas ;) (performance is better) 

I have done test, on my test cluster (3 servers with simple raid1 with baterry cache), 
it take around 10s to create a 500GB qcow2 file with preallocate metadatas. 

it take 0,5s to create 500qcow2 without preallocate. 

(But I have tested with qemu 1.7) 


Do you have problem only on backup restore ? or also when you create a new disk ? 


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

De: "Ferrara" < aixsyd at gmail.com > 
À: "Alexandre DERUMIER" < aderumier at odiso.com > 
Envoyé: Mercredi 8 Janvier 2014 03:51:39 
Objet: Re: gluster patch 



I think I may have found a solution. Its shockingly simple. Theres a line that reads: 


push @$cmd, '-o', 'preallocation=metadata' if $fmt eq 'qcow2'; 



I changed it to: 


# push @$cmd, '-o', 'preallocation=metadata' if $fmt eq 'qcow2'; 



And it seems to have worked for VM creation. the qcow2 image it made is 194KB. I'm going to try to boot/install to it. I'll letcha know what I find. 


Sam 



On Tue, Jan 7, 2014 at 9:27 PM, Ferrara < aixsyd at gmail.com > wrote: 




Not sure what I did, but I tried it again, and it seems to be doing the same thing - lock timeout, aborting. It didnt throw the previous error. 



Sam 





On Tue, Jan 7, 2014 at 9:23 PM, Ferrara < aixsyd at gmail.com > wrote: 

<blockquote> 


Restoring from a snapshot warrants the following: 


Parameter verification failed. (400) 

storage : invalid format - storage ID '' contains illegal characters 






On Tue, Jan 7, 2014 at 9:10 PM, Alexandre DERUMIER < aderumier at odiso.com > wrote: 

<blockquote> 

Hi, 

Here the glusterplugin with patch include. 

just replace /usr/share/perl5/PVE/Storage/GlusterfsPlugin.pm 

and restart 

/etc/init.d/pvedaemon restart 
/etc/init.d/pveproxy restart 
/etc/init.d/pvestatd restart 



(The plugin create the volume with qemu-img gluster:// , instead using fuse mount point. 

So, this apply also on backup restore, when the file is created. 

Tell me if it's works better for you. 








 

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

</blockquote>


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



More information about the pve-devel mailing list