incubator-cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marcus Sorensen <>
Subject [DISCUSS] resizeVolume
Date Tue, 08 Jan 2013 18:55:45 GMT
   I've pushed this as a new branch called 'resizevolume', in order to
collaborate with a few of you on it. I have several questions on how to
proceed, due to refactoring efforts. I'm hoping that the experts in those
efforts can collaborate.

1) Wido, would you be interested in looking at adding RBD resize to
LibvirtComputingResource? The resize for CLVM and QCOW2 is done via
scripts/storage/qcow2/ at the moment, since the libvirt
bindings won't support volBlockResize until you get 0.5.0 out (and even
then I'm not sure if it will support anything more than just informing the
hypervisor/guest of the change, which is also in the script).

2) I'm not sure exactly how this will be affected by the storage refactor,
but I'm sure it will be. I looked for examples in the createVolume on
javelin, but it seems that the storage refactor is still early on, and per
Edison's update only works for a few template commands at the moment. I
thought it best to get this in as a fully fleged member of the storage API
so that when the work is done on the storage commands, this will get looked
at as well. So if everything looks good otherwise, I'd like to not block on
the storage refactor unless this is just going to be scrapped when that
comes along.

3) I've implemented resize for Xen, but only tested with local storage SR.
The SR in devcloud doesn't seem to support online resize, and I'm not
familiar enough with the Xen setup to really figure out if that's just a
byproduct of how the SR was created, whether it's a Xen limitation in
DevCloud, or what. If someone more familiar with Xen wants to take a peek
it would be appreciated, however I'm not really considering it a blocker if
nobody does. For now resize works offline (or with data disk detached) and
fails with warning that VM needs to be stopped or data disk detached.

4) Can someone who is working with the api refactoring take this into
account? I think I'd prefer it go in before the refactor, or if it doesn't
that someone help me make any modifications necessary, so that it's not
broken at any point in master.

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message