Recent blog entries for amits

24 Nov 2014 (updated 24 Nov 2014 at 11:27 UTC) »

Upgrading to Fedora 21

The Fedora Project will soon put out its 21st release.  I’ve been running the pre-release bits for a while now, here are a few observations:

  • Upgrade from Fedora 20 to Fedora 21 via ‘fedup‘ was fast on my SSD disk, and there were no blockers after the reboot – minimal downtime!
  • Bug 740607 – evince no longer can switch to prev / next pages using the buttons or the ctrl+up/down keyboard shortcuts
  • Bug 740608 – gnome-shell’s calendar display overflows from the box if the number of calendar entries are more than some number; the box is always fixed in size.
  • Bug 739991 / Bug 730128 – gnome-terminal doesn’t pass alt+<n> to applications running inside the terminal if there isn’t a tab with that <n>.  This is the most serious regression for me; breaks several workflows for me: my irssi session as well as non-irssi terminals I use for work.  A surprising thing I found out after filing this report is there’s no way to open a closed bug report on gnome bugzilla, which means if some decides the bug isn’t going to be fixed, there’s no option to get new information back on the developers’ radar.
  • Bug 1163747 – memleak in upowerd

The workarounds[1] listed[2] earlier[3] are still in effect for things to work to my liking.

Everything else seems to be working reasonably fine so far, no further regressions.  I am tempted to give KDE a try again, though!

Syndicated 2014-11-24 05:39:33 (Updated 2014-11-24 10:32:55) from Think. Debate. Innovate.

My talk at the CentOS Dojo Pune 2014

I spoke at the CentOS Dojo in Pune yesterday on new features available in CentOS release 7.0 since the 6 release.  Slides are available here: What’s New in Virtualization.  The event was organized by the Pune GNU/Linux Users Group (PLUG) for the CentOS project.

My talk was scheduled as the last talk of the day.  I was already quite tired by the time the talk started, and was totally exhausted when it finished.

There were about 30 people attending, with some of them having already used KVM.  There were quite a few questions related to KVM and how it compares to other hypervisors, and about features supported by KVM.  I was happy with the interaction, as well as the questions I received.  It showed a nice interest towards virtualization and KVM.

Also nice to see that some were using virt-manager, oVirt, etc., already.  I couldn’t always answer everything related to the higher levels, but pointed people at bugzilla for bugs and the mailing lists for questions.

Syndicated 2014-11-23 09:02:49 from Think. Debate. Innovate.

Speaking at the CentOS Dojo, Pune

I’ll be speaking about KVM, progress since EL6, and other virt stuff at the CentOS Dojo in Pune this Saturday, 22nd November.  If you’re in Pune, feel free to register and drop by!

Syndicated 2014-11-17 09:56:30 from Think. Debate. Innovate.

Fedora Activity Day: Security I

Last Saturday a few of us gathered to work on Fedora Security.  This FAD (Fedora Activity Day) was the second in recent times held in Pune, after the testing FAD held in August.

Security FAD

The goal of the FAD was to get introduced to the newly-formed Fedora Security Team, pick up a few bug reports that were tagged as security-relevant bug reports, and triage them.  Fixing the bugs wasn’t part of the agenda, as actually pushing package updates needs one to be a provenpackager or the maintainer of the package.

We were assembled at the Red Hat Pune office.  I took a shot at transcribing PJP’s intro talk on the #fedora-india IRC channel, and a couple of people joined remotely in the triaging activity, which was quite nice to see.

The FAD wiki page had all the relevant information on how to go about triaging the bugs, so it was all quite straightforward from there.

I got a bit bored by just going through bug reports, without much “action” happening — it depended on the bug we selected on whether we just needed to set needinfo? on the assignee of the bug, or actually check progress of packages upstream, whether a patch was available, etc.  I just looked through bugs which looked relevant to virtualization, and then wanted to look at different ways to contribute.

PJP suggested looking at some fuzzers, and actually running them.  He pointed me to Radamsa as an example.  That does look like a good tool to generate some random input to programs, and see how they behave under unexpected input.  I didn’t actually get to run it, but now have an idea on what to do when I feel bored again.

While reading about Radamsa, I also thought a bit on how to fuzz qemu.  Nothing concrete came up, but one thought is to send weird stuff from guests to the host, by way of weirdly-formatted network packets (to test virtio-net or other net device emulations), or block device requests (to test virtio-blk / virtio-scsi / ide / ahci).  That’s an idea for a side project.

There also was a Docker meetup running at the same time at the office, so I dropped in there a couple of times to see what they were upto.  The organizers had split the session into talks + hackathon; and both were very well-attended.  In my lurking there, I overheard what Kubernetes is about, and a few terminologies it introduces into the tech world: minions and pods.  I’m sure we’re going to run out of words in the English language to re-purpose to technical usage very soon.

The FAD was originally supposed to happen in September, but got delayed to November.  For the next installation of Fedora-related activities, we may do an F21 release party along with a few user talks.  Regular FADs should resume in January, I suppose.

Syndicated 2014-11-03 18:28:12 (Updated 2014-11-03 18:29:12) from Think. Debate. Innovate.

KVM Forum 2014

It’s been a couple of weeks that I’ve returned from Düsseldorf, Germany, after attending the seventh KVM Forum; an event where developers and users of the Linux virtualization technology gather to discuss the state of the hypervisor and tools around it, and brainstorm on future plans. As with the previous few years, the event was co-located with LinuxCon Europe.

IMA_4326

A few observations from the event, in random order:

  • Linux Foundation did a great job of hosting and planning the event.
  • This was the first time when the food was great!  There were even options for vegetarians, vegans, and kosher food.
  • The venue, Congress Centre Düsseldorf, was huge, and located perfectly along the picturesque Rhine river.
  • It was the first KVM Forum which Avi did not attend.
  • The schedule was nicely-paced, with not too many parallel talks, and plenty of opportunities for hallway discussions and meeting people.
  • Co-locating with the Linux Plumbers Conference, LinuxCon, CloudOpen, etc., conferences ensured there were a lot of people interested in Linux in general; and since almost everyone is at least a user of virt technologies, discussions with almost anyone is fruitful around how KVM/QEMU/libvirt get used, and what users expect from us.
  • All the talks were recorded on video, and are available in this youtube playlist.
  • Photos from the event are here
  • All the slides from talks are at the KVM Forum wiki page
  • The QEMU Summit was also held along with the Forum; notes from the Summit are posted on the qemu devel mail list.
  • Jeff Cody’s talk on an intro to writing and submitting patches to qemu, and working with the community, got very positive feedback.  At least two people told me it would’ve been good to have that talk a year back, when they were getting started.  Well, it’s now available on the ‘net, and archived for people just starting out!
  • The OVA (Open Virtualizaiton Alliance) session on connecting users and developers of KVM by hosting a panel of KVM users (from cloud providers / builders) had one interesting insight: everyone wants more performance from KVM networking (well, the session was focussed on NFV, so that isn’t surprising).  No matter how fast you go, you want stuff to go faster still.  No one talked of stability, reliability, manageability, etc., so I suppose they’re just happy with those aspects.
  • A discussion with Chris Wright on KVM and OpenStack brought to me a surprise: KVM “just works” on OpenStack, and KVM is not the layer where there are problems.  No matter how many features we add or how much more performance we can eke out of the hypervisor, the most user-visible changes are now going to happen in the upper layers, most specially within the OpenStack project.  Obviously, there’s a need for us to collaborate with the OpenStack teams, but for most purposes, KVM is hardly the bottleneck or blocker for taking stuff to the clouds.  (We do have a huge list of things to do, but we’re ahead of the curve — what we are planning to do is needed and anticipated, but we need a better way to expose what we already have, and the OpenStack teams are going full-throttle at it.)

I suppose these are the highlights; I may have forgotten a few things due to the intervening holiday season.

Syndicated 2014-11-03 17:59:47 from Think. Debate. Innovate.

KVM Forum 2014 Schedule

The 2014 edition of KVM Forum is less than a week away.  The schedule of the talks is available at this location.  Use this link to add the schedule to your calendar.  A few slides have already been uploaded for some of the talks.

As with last year, we’ll live-stream and record all talks, keep an eye on the wiki page for details.

One notable observation about the schedule is that it’s much relaxed from the last few years, and there are far fewer talks in parallel this time around.  There’s a lot of time for interaction / networking / socializing.  If you’re in Dusseldorf next week, please come by and say ‘hello!’

Syndicated 2014-10-09 19:34:42 (Updated 2014-10-09 19:51:08) from Think. Debate. Innovate.

OpenStack Pune Meetup

I participated in the OpenStack Meetup at the Red Hat Pune office a few weekends ago.  I have been too caught up on the lower-level KVM/QEMU layers of the virt stack, and know there aren’t too many people involved in those layers in Pune (or even India); and was curious to learn more about OpenStack and also find out more about the OpenStack community in Pune.  The event was on a Saturday, which means sacrificing one day of rest and relaxation – but I went along because curiousity got the better of me.

This was a small, informal event where we had a few talks and several hallway discussions.  Praveen has already blogged about his experiences, here are my notes about the meetup.

There were a few scheduled talks for the day; speakers nominated themselves on the meetup page and the event organizers allotted slots for them.  The proceedings started off with configuring and setting up OpenStack via DevStack.  I wished (for the audience present there) there would’ve been an introductory talk before a deep-dive into DevStack.  I could spot a few newbies in the crowd, and they would have benefitted by an intro.

In a few discussions with the organizers, I learnt one of their pain points for such meetups: there inevitably are newbies at each meetup, and they can’t move on to advanced topics just because they have to start from scratch for each meetup.  I suggested they have a clear focus for each meetup: tell explicitly what each meetup is about, and the expertise level that’s going to be assumed.  For example, there’s nothing wrong with a newbie-focused event; but then some other event could focus on the networking part of OpenStack, and they assume people are familiar with configuring and deploying openstack and are familiar with basic networking priciples.  This suggestion is based on the Pune FADs we want to conduct and have in the pipeline; and was welcomed by the organizers.

Other talks followed; and I noticed a trend: not many people understood, or even knew about, the lower layers that make up the infrastructure beneath OpenStack.  I asked the organizers if they could spare 10 mins for me to provide a peek into the lower levels, and they agreed.  Right after a short working-lunch break, I took the stage.

I spoke about Linux, KVM and QEMU; dove into details of how each of them co-operate and how libvirt drives the interactions between the upper layers and the lower layers.  Also spoke a little about the alternative hypervisor support that libvirt has, but the advantages of the default hypervisor, QEMU/KVM has over others.  I then spoke about how improvements in Linux in general (e.g. the memory management layer) benefits the thousands of people running Linux, the thousands people running the KVM hypervisor, and in effect, benefit all the OpenStack deployments.  I then mentioned a bit about how features flow from upstream into distributions, and how all the advantages trickle down naturally, without anyone having to bother about particular parts of the infrastructure.

The short talk was well received, and judging by the questions I got asked, it was apparent that some people didn’t know the dynamics involved, and the way I presented it was very helpful to them and they wanted to learn more.  I also got asked a few hypervisor comparison questions.  I had to cut the interaction because I easily overflowed the 15 mins allotted to me, and asked people to follow up with me later, which several did.

One of the results of all those conversations was that I got volunteered to do more in-depth talks on the topic at future meetups.  The organizers lamented there’s a dearth of such talks and subject-matter experts; and many meetups generally end up being just talks from people who have read or heard about things rather than real users or implementers of the technology.  They said they would like to have more people from Red Hat talking about the work we do upstream and all the contributions we make.  I’m just glad our contributions are noticed :-)

Another related topic that came up during discussions with the organizers are hackathons, and getting people to contribute and actually do stuff.  I expect a hackathon to be proposed soon.

I had a very interesting conversation with Sajid, one of the organizers.  He mentioned Reliance Jio are setting up data centres across India, and are going to launch cloud computing services with their 4G rollout.  Their entire infrastructure is based on OpenStack.

There were other conversations as well, but I’ll perhaps talk about them in other posts.

Internally at Red Hat, we had a few discussions on how to improve our organization for such events (even though they’re community events; we should be geared up to serve the attendees better).  Mostly included stuff around making it easier to get people in (ie working with security), getting the AV equipment in place, etc.  All of this was working fine during this event, but basically ensuring all of the things that do go right are also part of the list of things to look at while organizing events so we don’t slip up.

Syndicated 2014-10-05 07:09:10 from Think. Debate. Innovate.

KVM Forum 2014

The KVM Forums are a great way to learn and talk about the future of KVM virtualization. The KVM Forum has been co-located with the Linux Foundation’s LinuxCon events for the past several years, and this year too will be held along with LinuxCon EU in Dusseldorf, Germany.

The KVM Forums also are a great documentation resource on several features, and the slides and videos from the past KVM Forums are freely available online. This year’s Forum will be no different, and we’ll have all the material on the KVM wiki.

Syndicated 2014-09-29 07:39:55 from Think. Debate. Innovate.

Planet Virt

For a long time various people have been telling me there’s not much information on the low-level / plumbing details of the virt stack on Linux. Especially information related to qemu and its various settings, devices, and so on.

Documentation surely is difficult to come by, but a quick and straightforward solution is to syndicate all of the blog posts that people doing virt development write into a common stream: a planet virt. I started hosting and testing such an instance on openshift, but was quickly pointed to the existing Virt Tools Planet by Rich Jones and Dan Berrange. Dan added the list of people whose blogs I followed for virt development to that instance.

I updated the KVM and QEMU wikis to ensure the Planet gets more visibility, and hope this goes a small way to quell the complaints of not enough available information.

Syndicated 2014-09-29 07:34:10 (Updated 2014-09-29 08:09:54) from Think. Debate. Innovate.

24 Aug 2014 (updated 29 Sep 2014 at 08:13 UTC) »

Fedora Activity Day Pune Report

I participated in the Fedora Activity Day at the RH office in Pune yesterday. There was a decent turnout, 20+ people, and it was fun to test the in-progress version of the upcoming F21 release along with other folks.

Siddhesh came up with the idea of rebooting Fedora-related activities in Pune, and a few of us showed interest in such an activity. We quickly agreed on what to focus on for the first such activity: test the upcoming release. This would give us an opportunity to improve the experience with F21, and also be a low-barrier-to-entry activity for first-time contributors: we have had some FADs in the past, but the people who turn up tend to be usually familiar with Fedora or particular aspects of the OS; so focussing on using the OS, and filing bugs along the way, was thought to be a great way to initiate newcomers without necessarily diving deep into technical details.

FAD Photo 3
FAD Photo 3

In that respect, I’d like to think the FAD was a success. We had people testing the installer, the GNOME and KDE desktops within VMs and via live images on laptops, and also a few specific items like VM snapshots and DNSSEC.

Before the FAD, I downloaded and tested in a VM two nightly images – the default workstation image with the GNOME desktop, and the KDE spin. The Aug 20 nightly for both the images worked fine, so we declared them as gold images for the FAD. Most people already had downloaded them before they came for the FAD, and this helped us start with the FAD as soon as our laptops were booted.

We started off at about 9 AM, and I was around till a bit after 4 PM. I tested the default GNOME live image on an X200 laptop, and also inside a VM. I found a couple of suspend-to-ram related issues in GNOME and in the kernel. Quite a few people tested the installer, and I liked how we kept conversing about the issues people were seeing, others attempting to replicate the issues, and once there were 2 or 3 +1s for a particular kind of an issue, we knew it was a fairly reproducible bug.

FAD Photo
FAD Photo 2

Some people ran through some Fedora Test Day test cases, some went through ON_QA bugs to provide karma on bodhi.

PJP took the lead on educating us on DNSSEC and walking us through setting it up as well as testing whether everything works fine.

Kashyap then spoke a bit about how VM snapshots are a great tool for testing destructive things, and showed the couple of different snapshot techniques, and how to set them up and use them using libvirt. Quite a few people agreed this was really cool, and I expect them to start using snapshots in their regular $DAYJOB activities.

We kept recording our progress, bugs found, etc., on an etherpad. The #fedora-india channel helped us exchange links and was helpful for general chit-chat. (Edit: The contents of the etherpad as on the day of the FAD are archived here)

Through the day, quite a few categories and components were tested, as noted in the etherpad.

The FAD wiki page‘s status area is still being populated, but the etherpad has links to bugs found and filed.

The initial mails about hosting a FAD to the fedora-india list showed there was interest in a lot of topics to cover for FADs, so I’m sure we’ll have more such FADs organized with more topics in the future.

Since this was mostly a volunteer-driven event, everything was quite informal. Red Hat sponsored the venue and snacks, and everything else was handled by us on the fly, like lunch being ordered when people started feeling hungry — according to preferences by show-of-hands. Since this FAD was cobbled together in very quick time, we didn’t have enough time to engage with the Fedora contacts for budget for swag or food; hopefully we will be able to get that sorted out soon — especially since we have interest in the FADs and also topics lined up to work on.

Syndicated 2014-08-24 08:50:40 (Updated 2014-09-29 07:42:12) from Think. Debate. Innovate.

84 older entries...

New Advogato Features

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!