On 03/19/2012 07:40 AM, Peter Krempa wrote:
On 03/17/2012 05:33 PM, Eric Blake wrote:
Offline internal snapshots can be rolled back with just a little bit of refactoring, meaning that we are now automatically atomic.
* src/qemu/qemu_domain.c (qemuDomainSnapshotForEachQcow2): Move guts... (qemuDomainSnapshotForEachQcow2Raw): ...to new helper, to allow rollbacks. * src/qemu/qemu_driver.c (qemuDomainSnapshotCreateXML): Offline snapshots are now atomic. --- You rollback changes to disks if other disks are not snapshotable, but later on, when the actual qemu-img command is run and fails the rollback is not performed.
Good catch.
I's suggest squashing in:
if (virRun(qemuimgarg, NULL) < 0) { if (try_all) { VIR_WARN("skipping snapshot action on %s", def->disks[i]->dst); skipped = true; continue; + } else if (STREQ(op, "-c") && i) { + /* We must roll back partial creation by deleting + * all earlier snapshots. */ + qemuDomainSnapshotForEachQcow2Raw(driver, def, name, + "-d", false, i); }
Yep, that looks right. Thanks for the test case.
Otherwise looks good. ACK with that suggested change.
Same story as for 3/3 - I'll wait to push this until I've run a few more tests for the rest of my pending series, in case I find any more last-minute issues. -- Eric Blake eblake@redhat.com +1-919-301-3266 Libvirt virtualization library http://libvirt.org