[zfs-discuss] R720xd + ZoL as iSCSI server

Cédric Lemarchand cedric.lemarchand at ixblue.com
Fri May 17 17:21:07 EDT 2013


Vmware iscsi software adaptater can only use a 512b LUN, wich is the default bs of most iscsi target (iet, scst, lio). 
VMFS 5 default to 1M bs. 
I think that for this uses (iscsi + vmware), a bs of 128k for the zfs dataset holding the LUN (sparse files) is good, because there are only big files  , ans compression is more efficient with large bs. 
Please correct me if I am wrong. 


--
Cédric Lemarchand

Le 17 mai 2013 à 22:13, "Dan Swartzendruber" <dswartz at druber.com> a écrit :

>> When generating the file:
>> 
>> dd if=/dev/zero of=/data/export_iscsi/volume01 bs=XXXXXX
>> 
>> Any ideal number for that? (if it matters at all)
>> 
>> 
>> On Fri, May 17, 2013 at 5:02 PM, Dan Swartzendruber
>> <dswartz at druber.com>wrote:
>> 
>>>> So this may be a little off-topic, but if I go for sparse files, is
>>> there
>>>> an ideal block size I want to give them for VMware to use?
>>> 
>>> Do you mean an iSCSI block size?
> 
> eeek, why do you want to pre-create the entire file?  if that's really
> what you want to do okay, otherwise just do this:
> 
> truncate -s ???G /data/export_iscsi/volume01
> 
> Apart from that, I'm not sure I get your point.  The above is for
> allocating and zeroing the LUN file - it hardly matters what block size
> you give dd to write the file - the client certainly won't see that...
> 


More information about the zfs-discuss mailing list