Following are the notes that I've taking during the call. I'm pretty sure that I'm missing something important, such as another blocker BZ, so please reply with corrections/addtions.
- Adam: going to introduce Python binding for the new Vdsm API
- Sharad has joined our call for the first time. Welcome! Interested in backup/restore functionality in oVirt, to facilitate integration with TSM, presumably http://www-142.ibm.com/software/products/us/en/tivostormana/
- Federico warns that live snapshot is not enough, since we lack live merge. After a year of daily backups, each qcow chain would be 365 hops deep.
ovirt-3.2: - We should probably revert http://gerrit.ovirt.org/9315 (integrate zombie reaper in supervdsmServer) as it tickles a bug in multiprocessing. See http://lists.ovirt.org/pipermail/users/2013-January/011857.html for more details in the thread starting in http://lists.ovirt.org/pipermail/users/2013-January/011747.html (thanks for raising the issue and finding the culprit, Dead Horse)
- Danken believes that all important el6-specific issues reported by dreyou are now handled in the master and ovirt-3.2 branches. Please report this list if he's wrong.
- Federico reports that udev is going to revert to its Fedora 18alpha behavior so that we can keep our integration with it. I'd love to see the udev BZ listed in our tracker https://bugzilla.redhat.com/showdependencytree.cgi?id=881006&hide_resolv... as I do not recall it. Federico, could you add it? We should probably require a newer udev release.
- Federico reports that we must take a short-but-important patch from Lee Yarwood: http://gerrit.ovirt.org/#/c/11281/
Cross-distro: - Adam: a guy from IBM is now listing issues that block running Vdsm on Ubuntu. Yay! - Saggi suggests that IBM contribute an Ubuntu slave for Jenkins, so that each and every patch is tested not to introduce Ubuntu regressions - just as Red Hat has recently done with EL6. - Danken: yes, we have got el6-based unit-testing running. If you are still getting an unjustified X, you may need to rebase on current master. - Federico: NetworkManager 9 has bridge support, which leads the way to NM-based implementation of Vdsm's configNetwork. - Danken: Toni is working on a suggestion for refactoring configNetwork in such a way that multiple implementations (e.g. ifcfg-based, NM-based) can coexist. Stay stuned for his Wiki page on the subject.
Other refactoring going on: - Vinzenz has begun breaking the horrible libvirtvm+vm to edible-size pieces. Reviews are most welcome, particularly from Mark Wu, whose http://gerrit.ovirt.org/10054/ only awaits verification tick.
Functional Testing: - Adam: a guy from IBM (Zhou, I presume) is running the functional tests via Jenkins on his laptop. Yay! - Adam: plans to add a functional test for the getVdsCaps verb, as he found out that an evil Vdsm developer has added values to the caps without updating the schema.
- P.S. FlowID is now in for storage verbs as well (thanks, Douglas). I hope Haim is happier now.
Happy coding!
于 2013-1-29 4:00, Dan Kenigsberg 写道:
Following are the notes that I've taking during the call. I'm pretty sure that I'm missing something important, such as another blocker BZ, so please reply with corrections/addtions.
Adam: going to introduce Python binding for the new Vdsm API
Sharad has joined our call for the first time. Welcome! Interested in backup/restore functionality in oVirt, to facilitate integration with TSM, presumably http://www-142.ibm.com/software/products/us/en/tivostormana/
Federico warns that live snapshot is not enough, since we lack live merge. After a year of daily backups, each qcow chain would be 365 hops deep.
That's horrible. It looks like that Qemu 1.3 and libvirt 1.0.0 already supports the "live block commit" which is the base of live commit. Do we have plan to support "live commit" in next oVirt release?
ovirt-3.2:
We should probably revert http://gerrit.ovirt.org/9315 (integrate zombie reaper in supervdsmServer) as it tickles a bug in multiprocessing. See http://lists.ovirt.org/pipermail/users/2013-January/011857.html for more details in the thread starting in http://lists.ovirt.org/pipermail/users/2013-January/011747.html (thanks for raising the issue and finding the culprit, Dead Horse)
Danken believes that all important el6-specific issues reported by dreyou are now handled in the master and ovirt-3.2 branches. Please report this list if he's wrong.
Federico reports that udev is going to revert to its Fedora 18alpha behavior so that we can keep our integration with it. I'd love to see the udev BZ listed in our tracker https://bugzilla.redhat.com/showdependencytree.cgi?id=881006&hide_resolv... as I do not recall it. Federico, could you add it? We should probably require a newer udev release.
Federico reports that we must take a short-but-important patch from Lee Yarwood: http://gerrit.ovirt.org/#/c/11281/
Cross-distro:
- Adam: a guy from IBM is now listing issues that block running Vdsm on Ubuntu. Yay!
- Saggi suggests that IBM contribute an Ubuntu slave for Jenkins, so that each and every patch is tested not to introduce Ubuntu regressions - just as Red Hat has recently done with EL6.
- Danken: yes, we have got el6-based unit-testing running. If you are still getting an unjustified X, you may need to rebase on current master.
- Federico: NetworkManager 9 has bridge support, which leads the way to NM-based implementation of Vdsm's configNetwork.
- Danken: Toni is working on a suggestion for refactoring configNetwork in such a way that multiple implementations (e.g. ifcfg-based, NM-based) can coexist. Stay stuned for his Wiki page on the subject.
Other refactoring going on:
- Vinzenz has begun breaking the horrible libvirtvm+vm to edible-size pieces. Reviews are most welcome, particularly from Mark Wu, whose http://gerrit.ovirt.org/10054/ only awaits verification tick.
Functional Testing:
- Adam: a guy from IBM (Zhou, I presume) is running the functional tests via Jenkins on his laptop. Yay!
Zhou Zheng Sheng zhshzhou@linux.vnet.ibm.com
Adam: plans to add a functional test for the getVdsCaps verb, as he found out that an evil Vdsm developer has added values to the caps without updating the schema.
P.S. FlowID is now in for storage verbs as well (thanks, Douglas). I hope Haim is happier now.
Happy coding! _______________________________________________ vdsm-devel mailing list vdsm-devel@lists.fedorahosted.org https://lists.fedorahosted.org/mailman/listinfo/vdsm-devel
On Mon, Jan 28, 2013 at 10:00:27PM +0200, Dan Kenigsberg wrote:
- Federico reports that udev is going to revert to its Fedora 18alpha behavior so that we can keep our integration with it. I'd love to see the udev BZ listed in our tracker https://bugzilla.redhat.com/showdependencytree.cgi?id=881006&hide_resolv... as I do not recall it. Federico, could you add it? We should probably require a newer udev release.
I've found it and posted http://gerrit.ovirt.org/11489 to require the systemd version that fixes https://bugzilla.redhat.com/903716.
On 01/28/2013 03:00 PM, Dan Kenigsberg wrote:
Following are the notes that I've taking during the call. I'm pretty sure that I'm missing something important, such as another blocker BZ, so please reply with corrections/addtions.
Adam: going to introduce Python binding for the new Vdsm API
Sharad has joined our call for the first time. Welcome! Interested in backup/restore functionality in oVirt, to facilitate integration with TSM, presumably http://www-142.ibm.com/software/products/us/en/tivostormana/
Federico warns that live snapshot is not enough, since we lack live merge. After a year of daily backups, each qcow chain would be 365 hops deep.
ovirt-3.2:
We should probably revert http://gerrit.ovirt.org/9315 (integrate zombie reaper in supervdsmServer) as it tickles a bug in multiprocessing. See http://lists.ovirt.org/pipermail/users/2013-January/011857.html for more details in the thread starting in http://lists.ovirt.org/pipermail/users/2013-January/011747.html (thanks for raising the issue and finding the culprit, Dead Horse)
Danken believes that all important el6-specific issues reported by dreyou are now handled in the master and ovirt-3.2 branches. Please report this list if he's wrong.
I could deploy CentOS 6.3 without any problem, host became UP. Since there is no EL6.4 repo available at moment, I have created few RPM updated packages required by VDSM: http://dougsland.fedorapeople.org/oVirt3.2-EL6/
Additionally is required EPEL + oVirt Nightly repo.
[ovirt-nightly] name=Nightly builds of the oVirt project baseurl=http://ovirt.org/releases/nightly/rpm/EL/$releasever/ enabled=1 skip_if_unavailable=1 gpgcheck=0
Federico reports that udev is going to revert to its Fedora 18alpha behavior so that we can keep our integration with it. I'd love to see the udev BZ listed in our tracker https://bugzilla.redhat.com/showdependencytree.cgi?id=881006&hide_resolv... as I do not recall it. Federico, could you add it? We should probably require a newer udev release.
Federico reports that we must take a short-but-important patch from Lee Yarwood: http://gerrit.ovirt.org/#/c/11281/
Cross-distro:
- Adam: a guy from IBM is now listing issues that block running Vdsm on Ubuntu. Yay!
- Saggi suggests that IBM contribute an Ubuntu slave for Jenkins, so that each and every patch is tested not to introduce Ubuntu regressions - just as Red Hat has recently done with EL6.
- Danken: yes, we have got el6-based unit-testing running. If you are still getting an unjustified X, you may need to rebase on current master.
- Federico: NetworkManager 9 has bridge support, which leads the way to NM-based implementation of Vdsm's configNetwork.
- Danken: Toni is working on a suggestion for refactoring configNetwork in such a way that multiple implementations (e.g. ifcfg-based, NM-based) can coexist. Stay stuned for his Wiki page on the subject.
Other refactoring going on:
- Vinzenz has begun breaking the horrible libvirtvm+vm to edible-size pieces. Reviews are most welcome, particularly from Mark Wu, whose http://gerrit.ovirt.org/10054/ only awaits verification tick.
Functional Testing:
Adam: a guy from IBM (Zhou, I presume) is running the functional tests via Jenkins on his laptop. Yay!
Adam: plans to add a functional test for the getVdsCaps verb, as he found out that an evil Vdsm developer has added values to the caps without updating the schema.
P.S. FlowID is now in for storage verbs as well (thanks, Douglas). I hope Haim is happier now.
Happy coding! _______________________________________________ vdsm-devel mailing list vdsm-devel@lists.fedorahosted.org https://lists.fedorahosted.org/mailman/listinfo/vdsm-devel
vdsm-devel@lists.stg.fedorahosted.org
-
Dan Kenigsberg -
Douglas Schilling Landgraf -
shuming＠linux.vnet.ibm.com