Interviews On My FOSDEM 2016 Talk
I was interviewed by the FOSDEM folks on my upcoming talk on KVM Live Migration.
I was also interviewed by Brian Proffitt, who has written an article that serves as a preview for the talk.
Looking forward to the talk this Sunday!
Syndicated 2016-01-27 22:09:33 from Think. Debate. Innovate.
SCaLE14x Talk: KVM Weather Report
I did a talk titled ‘KVM Weather Report‘ at the SCaLE conference in Pasadena, California yesterday. The slides are available at this location.
The talk introduced the KVM stack (Linux, KVM, QEMU, libvirt); briefly went over some features and the communities around the projects, and discussed some of the new features added to the KVM stack in the last year.
Next up is my talk on live migration of VMs at FOSDEM in Belgium.
Syndicated 2016-01-25 13:18:21 (Updated 2016-01-25 13:19:46) from Think. Debate. Innovate.
QEMU Maintainer Interviews for the 2.5 release
Hot on the heels of the QEMU 2.4 release, we have QEMU version 2.5 releasing today.
QEMU creates the virtual machine which guest operating systems run on top off. QEMU also handles host-specific things, like the storage and networking on the host.
Given the wide scope of this project, there are several changes that many contributors add to each release. To repeat the success with the 2.4 release video, I asked maintainers to record segments for the 2.5 release as well. A few maintainers and contributors chipped in with videos, and a few updated the ChangeLog page, and added new feature pages. Thanks to all who pitched in!
(Link to video on YouTube: https://www.youtube.com/watch?v=lFgopoa9Rso)
In the video, we have Michael Tsirkin speaking on new security features, virtio 1, vhost and vhost-user and some guest enhancements for PCI and PCI-E; Juan Quintela about live migration enhancements (like the new autoconverge implementation); Dave Gilbert about the new postcopy migration feature; Eric Blake on QMP introspection; Stefano Stabellini on Xen; Stefan Hajnoczi on enhancements in the block layer; and Alberto Garcia expanding on a few block layer enhancements.
Side note: if you notice, the new feature pages linked above are quite detailed. They’re based on a template which was prepared based on feedback from quite a few people, including Docs and Quality Engineering folks from Red Hat, the Fedora Changes template, and my experience with answering questions for my features. I included as many items and questions in the template as possible so it’s easier for users to get as much information on a feature from the feature page before they ask the developer for more information. In the coming days, I hope to point out the new template to more people, and perhaps make it a requirement for new features landing in QEMU.
As with the last time, I used OpenShot to create the video, and Audacity for some noise reduction on the audio track of one of the videos. Unfortunately, this time, OpenShot on Fedora 23 crashed a lot – approx. once for each action I performed. Enabling the autsave functionality helped a lot; I lost several edits before I did that 
Syndicated 2015-12-16 15:30:14 (Updated 2015-12-16 14:15:23) from Think. Debate. Innovate.
Let’s Encrypt on OpenShift
Let’s Encrypt have lauched their public beta, and they’re now offering SSL certificates to everyone. The process is very easy and quite easy to automate. However, there’s a catch: these certificates expire in a few days (90 days as of now), so they have to be renewed often. That’s where having the process be simple and automatable helps.
OpenShift doesn’t yet have a way to automate SSL certificate installs; so adding an SSL cert to this blog is going to be a manual process every few days. I’m on the OpenShift silver tier (not yet available in my region, but it’s a perk available to Red Hat employees), so I get to attach a custom cert to my site.
Here’s what I did to get an SSL cert for this blog and enable it: installed the letsencrypt package from the Fedora repos, and ran this command:
$ letsencrypt --text --email=<email-id> --domains log.amitshah.net --agree-tos --renew-by-default --manual certonly --config-dir ~/.letsencrypt/etc --work-dir ~/.letsencrypt/var-lib/ --logs-dir ~/.letsencrypt/var-log/
I had to verify that I actually own the log.amitshah.net domain, so I logged into the OpenShift instance and created a file there as instructed by the prompts during the above command.
This article on the Fedora Magazine has more detail on the letsencrypt command. I gave custom paths for the log, work, and config dirs, as I didn’t want to run the tool as root.
That’s it; I had the certs available in ~/.letsencrypt/etc/. I then went to the OpenShift Online web console and uploaded the files there, and SSL was instantly available on the site.
Syndicated 2015-12-16 11:25:51 (Updated 2015-12-16 11:35:11) from Think. Debate. Innovate.
30 Years of GNU and Software Freedom Day
It’s 30 years of GNU — 30 years of freedom and 30 years of owning one’s computers. I can’t imagine a life where I don’t have control over the software I run. I’m going to be eternally thankful to RMS and Linus for starting the mass movements that have not only transformed an entire industry, but also shaped my thinking and my career.
A few Red Hatters (including yours truly) have shared stories of their first brush with free software here — give it a read, it’s a good trip down the memory lane, as well as some inspiring anecdotes from people who have been involved with free software for a really long time.
Here’s wishing everyone a liberating Software Freedom Day (Sep 19th), and many more years of freedom to everyone!
Syndicated 2015-09-18 17:48:45 from Think. Debate. Innovate.
QEMU Maintainers on the 2.4 Release
QEMU is the software that creates virtual hardware which guest operating systems run on top of. All (well, almost all — see note below[*]) the hardware that a guest OS has access to is actually written to some specifications in software — i.e. no physical hardware is involved. For the QEMU/KVM hypervisor, most of these devices are written in the QEMU source repository. A few devices are part of the KVM code in the Linux kernel. QEMU also handles a lot of host-specific stuff, like storage and networking for the virtual machines.
[* Exception: physical hardware devices assigned to guests.]
Many contributors to the QEMU and KVM projects meet at the annual KVM Forum conference to talk about new features, new developments, what changed since the last conference, etc.
The QEMU project released version 2.4 just a week before the 2015 edition of KVM Forum. I thought that was a good opportunity to gather a few developers and maintainers, and get them on video where we can see them speak about the improvements they made in the 2.4 release, and what we can expect in the 2.5 release.
(direct link to video, also on QEMU’s G+ page)
In addition to the video, I had also asked a few QEMU maintainers via email before the KVM Forum on things they had merged for the 2.4 release. Some of the maintainers, who couldn’t be in the video, sent in updates:
Thanks to all the maintainers who replied to my emails and those who agreed to appear on video. I shot the videos on my phone, so the quality is quite low. I also didn’t realise the recording didn’t start till a couple of seconds after I pressed the record button, so the intro with the name of the maintainers in some videos are clipped. Next time should be better!
OpenShot was used to collate all the individual videos, and Audacity was used to mix in the background music, and remove background noise on some of the clips. Huge thanks to the teams developing these softwares as well!
Syndicated 2015-09-07 12:25:00 (Updated 2015-09-10 16:27:47) from Think. Debate. Innovate.
Going to KVM Forum 2015
In its 8th edition, the KVM Forum is moving back to North America this year, co-located with LinuxCon NA in Seattle. It starts with a KVM + Xen hackathon and the (invite-only) QEMU Summit on the 18th August, followed by talks and BOFs on the 19th, 20th, and the 21st. The schedule is here. To add all talks to your calendar, use this ics. The KVM Forum wiki page will have information on live streaming of talks, videos, slides, etc.
If you’re going to be around, please come up and say hi!
Syndicated 2015-08-04 07:00:58 (Updated 2015-12-15 09:26:06) from Think. Debate. Innovate.
Spot the difference
Yeah. The monkey’s sharper than Jared in the selfie dept.
Jared, thanks for being a friend of FUDCon India. You rock!
Syndicated 2015-07-11 18:12:55 from Think. Debate. Innovate.
Live Migrating QEMU-KVM Virtual Machines: Full Text
I’ve attempted to write down all I said while delivering my devconf.cz talk on Live Migrating QEMU-KVM Virtual Machines. The full text is on the Red Hat Developer Blog:
http://developerblog.redhat.com/2015/03/24/live-migrating-qemu-kvm-virtual-machines/
Syndicated 2015-03-24 15:53:40 (Updated 2015-05-31 07:55:13) from Think. Debate. Innovate.
FUDCon Pune 2015: CFP Closed, But…
The CFP for FUDCon Pune 2015 is now closed. We have had an overwhelming response: 141 talks/workshops submitted. This is more than twice the number of sessions we received for the 2011 edition. Talk submissions were pouring in till the last minute when we flipped the switch for the CFP page.
However, there’s a twist in the tale, so people who couldn’t get their CFP in aren’t left out.
For FUDCon Pune 2011, we tried something new: instead of the regular barcamp style at FUDCons, we put out a CFP and accepted talks in advance. This was done as we knew a FUDCon in India was going to be different than the usual FUDCons, where at most a hundred people turn up. Barcamp style is manageable with those numbers. We were proved right when the number of attendees exceeded 1000 then. Setting up the Drupal + COD instance back in 2011 had helped us with getting talk proposals accepted, as well as opening voting for everyone, so we knew which talks were going to be popular and schedule accordingly. Also, the voting aspect of it had a barcamp feel to it.
Fast forward to the 2015 edition. The conference is happening end of June, and our CFP closes start of March. That means there’s almost 4 months between the CFP close and the conference start. People who plan late, and have something to bring to the conference, are left out (well, there always are hallway tracks, but that restricts the audience). Also, just having the CFP isn’t traditional FUDCon style.
So I proposed in the 3rd March weekly FUDCon planning meeting that we do some barcamp tracks in addition to the regularly-scheduled tracks. We squabbled a bit over the details on how to actually make this work — have a new CFP open / have people submit talks via our COD instance, or do it on a whiteboard at the venue / do we recognise barcamp speakers as our scheduled speakers (matters for the swag!) / do we provide subsidy to barcamp speakers / etc.
One thing became clear from these discussions: no one objected to having the barcamp; we just had to flesh out the details. So here’s some good news for people who couldn’t submit their session for the CFP:
We’re going to have barcamp tracks at the FUDCon, and sessions can be proposed at the venue on the days of the conference.
One of the reasons to have a FUDCon is to get people together and work on stuff. The traditional FUDCon style (i.e. barcamps) were designed so that people active in Fedora just turned up, and things happened. It didn’t matter who got to speak officially; everyone just mingled together and had their say. “Hallway tracks” are always the more interesting and productive ones than the scheduled tracks. They’re also far more interactive and focussed.
We don’t want to take that away. This obviously means that whoever turns up to the event gets to propose barcamp sessions. So if people are thinking they’ll attend the event if their talk submission gets accepted, that’s the wrong way to go about it. Just turn up, you’ll have a voice no matter what.
This also means people needing travel sponsorship need to put in their subsidy requests NOW. We are soon going to start looking at requests already in the queue, so don’t be late for requesting subsidies.
Doing the barcamp in addition to formal scheduling is going to put a strain on the venue and logistics team, but we like challenges, and we’ll step up to them.
During the 10th March planning meeting, many people were of the view that we should not accept barcamp talk proposals on the COD instance. Reasons were that’s not how barcamp submissions work, and a whiteboard at the venue can be used just as well. That obviously means the barcamp tracks can only be held on the 2nd and 3rd days, since the first day people will have to come up, propose sessions, and vote on the proposed ones.
Since most of the people were of this view, we accepted this. However, we may change this depending on how our scheduling goes for the main conference (we can’t really fit in 140 talks!) so we may just flip over the rest to the barcamp track, and have people vote beforehand. I find there’s a certain appeal to using the COD instance to ease our jobs. It also means we have a proper schedule to display on the web / print out, so people know where to head to, instead of looking at the schedule on the web for the formal schedule, and missing out on the barcamps entirely, or having to hunt for the barcamp schedule at the venue. I’m sure we’ll have a lot more discussions around this topic as we approach the event.
The response to the CFP has been mind-boggling. We can project that the event itself will witness more participation than the 2011 edition. We did a fairly good job in 2011 handling 1000 people, 60 talks and 6 parallel sessions; this time we may have a much bigger scale, and we actually know what to expect; I’m confident we will execute it as well as the last time.
Update 16 Apr 2015: Corrected note about proposing barcamp submissions: can be proposed at the venue; instead of “till the day of the conference”.
Syndicated 2015-03-11 00:40:46 (Updated 2015-04-16 14:11:03) from Think. Debate. Innovate.
New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.
Keep up with the latest Advogato features by reading the Advogato status blog.
If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!
