iSCSI resize file extent

[MichelZ]

Hi all

I have not found a definitive answer to this yet, so I'm curious what the experts can tell me :)
Is it possible to resize an iSCSI file extent on-the-fly, without restarting the iSCSI daemon?

I have a 8.3.1-RELEASE box attached to an ESX 5.1 box using iSCSI.
iSCSI has LUC enabled with the default options & controller auth method "none"

When I change the size in the GUI, the GUI does update well.
the console log shows:

Apr 6 16:49:57 <storagename> istgt[26920]: reload configuration #5
Apr 6 16:49:57 <storagename> istgt[26920]: istgt_lu.c:2375:istgt_lu_reload_update: ***ERROR*** update active LU1: Name=iqn.2011-03.example.org.istgt:prod1, # of TSIH=4

and refreshing ESX does not see the new size / new free space.

What then puzzled me even more was that after a restart of the iSCSI service, ESX does still not see the new free space.

What am I doing wrong?

Thanks

 

[pdanders]

A better long term approach might be to use sparse files for your extents that are "magically large" (much larger than available disk) and then set quotas to match the actual current disk size.

That way you can easily expand when you grow your zpool

 

[jgreco]

You want to be really careful if you do something like that. vmfs takes great exception to it all if you try to write to disk and it cannot allocate space. You will want to be super-cautious with how you create your vmfs to make sure that it is living within existing space, which is best preallocated with a zero-filled file. So if you have 500GB, you should probably dd yourself a 300-400GB zero file (80% ZFS rule, 60% my iSCSI performance rule) and then you can trunc(1) the file longer if you wish, but when creating the vmfs on it, be darn sure to make it only as big as the zeroed space.

 

[pdanders]

You definitely need to know what you are doing, whenever I use that configuration I make extra, super sure the quota is less than the pool size.