hi all
hi
hi
sailus: ping
hverkuil: just created an account for stanimir.varbanov@linaro.org
sailus: started reviewing the report
I'm not sure what to do about https://patchwork.linuxtv.org/patch/52778/, since there no Signed-off-by.
you should ask for it
mchehab: ah, great. I'll check with Stan to see if I can redelegate all venus issues to him.
hmmm "This is more correct, but probably still wrong."
mchehab: the patch is a copy from
Enable Sony CXD2837ER slave demon on the Astrometa DVB-T2, known as the 2018 update.
sorry bad copy paste :/
the patch is a copy from https://github.com/torvalds/linux/pull/567, I should ask for a signed-off-by there. And the code needs some fixes
brb
hverkuil: I still need to send him the instructions
btw, patchwork is a different thing
I'm creating him an account for him to store stuff at linuxtv.org, not a patchwork account
Pong.
syoung: do your usual patch review, asking the guy to add his SOB to it
pointing to some place where the SOB definition is (we have it on our Wiki, for example)
back
syoung: you probably need to ask  "kapitanf" github user to SOB too, as he seems to be the original author of the first version
mchehab: he'll get patchwork access as well? That's perhaps the most important part of being a media maintainer.
mchehab: ok, thanks. There are some more dvb patches that need reviewing. I'm a bit slow since I have to read up a lot, so I don't think I'll have them all done before the next merge window.
hverkuil: let's go one step at the time
my main concern with patchwork is that it doesn't have any sort of changes log
syoung: the daily build has a few dvb-related warnings (mostly smatch I believe). It would be great if you could take a look at those.
anyone there can potentially disrupt the entire patch handling and we wouldn't know who
so, I prefer to be very conservative adding people there
mchehab: but that means you or me still would have to keep track of venus patches. That defeats the purpose.
mchehab: I can check this out.
sailus: the main issue is how to migrate the database
I was wondering updating to newer patchwork as well, we've been planning that for years...
mchehab: Do we need that?
People would just be required to use two patchwork instances for a while.
yes, at least for the patches that are tagged as new
I think moving the database could take more time than just that, unless there is a script provided to do the conversion.
I'm pretty sure that, if we migrate, we'll end by keeping stuff at the old instance
mchehab: What needs to be kept there?
If we could reduce the amount of patch to a manageable amount, that would be easier, I guess
We could leave it up and running for now, and shut it down until it's no  longe rneeded.
We could also tell people to resubmit, as other subsystems do.
They already do that for us, too...
Well, this doesn't need to be decided now.
Let's see how the new Patchwork version is first.
sailus: it shouldn't be hard to migrate the patches themselves
I can live with two patchwork instances.
with some script that would get each patch and submit at the second instance
but that will lose any discussions
as it won't get the followup e-mails
anyway, I'll seek for some time to download the patchwork database and run some tests
The new patchwork does have history of who did what?
hverkuil: patchwork gets all e-mails that refer to patches
adding them into a blog-like entry
Sorry, not what I meant.
Does the new patchwork log who changed the patch state (accepted, rejected, etc)?
on our version, no
I've no idea if this was fixed upstream
but I would prefer to move first to upstream, then find a way to implement it
and be sure that upstream get such patches first
with regards to patch handling, my plan is to review pending pull requests today and tomorrow, after finishing reviewing the media summit report
But to come back to patchwork rights: if we decide to trust someone to maintain his driver, but not to handle the patches in patchwork, then that defeats the whole purpose of having "trusted maintainers".
I hope to do more reviews tomorrow/Monday for longer patch series. And possible create pull requests if the code is OK.
hverkuil: two comments
1) I don't like the term "trusted" for maintainers/developers... as it implies that all the rest are "untrusted" :-)
It might be useful to be able to limit which patches a maintainer can touch in Patchwork. Again, I'm not sure if that's supported or not. Probably not.
Perhaps in the future someone could implement that. :-)
It'd be much easier to give maintainers the needed rights.
2) yeah, I understand your point. Ideally, they should be able to handle their own patches
1) a better name is welcome, but isn't that really what this is about? I.e., we trust them enough so that they can do their own patch review and make pull requests?
the best model would be if we could delegate them patches, and let them handle only the stuff delegated to them. Our patchwork version doesn't support it
it has an "all or nothing" model
actually, it has a more complex permission model, but last time I checked, there was no way to "partially" give patch changes to an user
the permissions are for the project as a hole
there was no way to give permissions to just the patches that a patchwork user's have
If maintainers have been behaving reasonable and have been doing a good job, then that should be enough. I mean, if that's not enough, then what other possible criteria would be needed?
hole -> whole :-)
hverkuil: that's why I'd say that we should start giving him an account
with time, as he earns "trustee points", we can give more
So will you keep track of venus patches in patchwork for him?
or mark them as "under review"
my scripts will mark them as accepted as we get patches merged
(we might create another patchwork status, but I think that this would be an overkill)
he'll need to update you (or me) for rejected patches
for any of us with patchwork rights to update it there
Well, you can explain that to him. I disagree with this policy (i.e. not giving him patchwork rights).
On another topic: I have one or two patchseries that I would like others to review. I'll post links here after the meeting (I need to dig them up).
I'm also working on creating a script that will use the virtual drivers and v4l2-compliance to do regression tests. Which is the reason why there are a lot of v4l2-compliance improvements being committed.
ad
oops, please ignore
good
ah, I forgot to mentionon our last meeting (I was really screwed due to TZ)...
I talked with LF about hosting our infra there
while in Vancouver
they'll see if it is possible to have an infra for us stored outside LF internal firewalls
(the way they currently work is that machines/VMs under their internal network have them as admin)
I'm waiting for some feedback if they can help us with that or not
mchehab: I'm not sure I like a dependency on LF that much...
the idea is that they won't be doing admin tasks inside the linuxtv VM, just at the host itself
I would prefer if we remained independent
mchehab: I remember there was some reservations during the Media summit towards LF hosting our services.
pinchartl: yes, I know you have some reservations about that
but the only way to be trully independent would be if we start a foundation ourselves
hosting a VM somewhere is easy and cheap. there is no need to depend on LF for that, there are plenty of options
We're now dependent on some German university and their goodwill in hosting us. Not something I like.
mchehab: I remember it was not pinchartl in particular who voiced those reservations.
hverkuil: I agree, that needs to be addressed.
sailus: I think it was a shared concern
For the record, I have no problem whatsoever with the LF hosting this.
pinchartl: I'd say the same. Not everyone probably thought that way, but I'd say many, if not most.
hverkuil: I agree we should move away from that, but if it's to switch from one dependency to another, I don't see the point :-)
sailus: I don't remember too many voices against that
If the LF disappears, then we have bigger problems.
in a matter of fact, I remember only pinchartl voicing it
From what I remember it was just pinchartl who argued against it.
it's not about the LF disappearing, it's about the control they would have
I agree with hverkuil: we're too dependent of LF
What I particularly remember was that the consensus was that the domain should be registered to "us".
if something gets wrong there, we'll have a lot more issues to concern
I'm pretty sure at least Gustavo shared my concern
what control? If they place it outside of their firewall then I assume we'd be admins anyway.
sailus: we're not talking about the domain or the admin rights
What was seen as a problem was not necessarily relying on LF on hosting, but eventually moving away from LF could pose a problem.
mchehab: I know, but it's related.
I'll keep those to us
mchehab: given how vocal you were against the Linux CoC which was pushed by the LF, I'm surprised you'd like to give them control over our community :-)
hverkuil: not the technical control, the political control
and the fact that moving away from the LF later could be seen as a hostile move
what do we need ? just a VM ? how much CPU/RAM/disk space ?
What I'd like to say is that finding a solution that everyone would be happy with would be better than choosing one with which only some are happy with.
sailus, pinchartl: I'm open to suggestions
mchehab: what are the technical needs ?
right now, just a VM is enough
do you know how much RAM and disk space ?
pinchartl: those are just technical details
As long as it is not a machine on someone's desk. I.e. it should be hosted by a reputable company (and I trust the LF more than most companies, to be honest. I don't understand the issues here). And how it should be funded.
pinchartl: I presume currently the git tree hosting is the largest resource consumer.
mchehab: Is that right?
mchehab: yes but they affect the price, that's why I'd like to have a rough idea
right now, the VM is not big
last time I checked, the traffic is high
can you check how big it is ? that would give us an initial estimate
pinchartl: anyway, the point right now is not a matter of money
that's just a small detail (as the technical details about the VM)
the point is that we need to trust on someone
I think the point is actually about money :-) if we had the budget, we wouldn't be discussing about it, would we ?
I share hans concerns
(10:55:07) hverkuil: As long as it is not a machine on someone's desk. I.e. it should be hosted by a reputable company (and I trust the LF more than most companies, to be honest. I don't understand the issues here). And how it should be funded.
pinchartl: if we had the money, we would still be discussing it
From the domain PoV, I've been researching a bit what that "us" could be.
mchehab: if we had the money we'd have a VM already, so we wouldn't be discussing the LF option
I could, for example, host it at Samsung...
I first thought we could find a foundation in Finland (we have two developers here), but the downside is that the languages are limited to Finnish and Swedish.
(or in the past at Red Hat)
So that is not an option.
We have contacts with the LF, they are already funding our media summits, we already rely on much of what they are doing, directly or indirectly.
but, IMHO, that would be a way worse than LF
You can't rely on companies like Samsung or Cisco or Intel to host things like this. You get a management change, and before you know it it is killed off.
sailus: why a random foundation in Finland would be any better than LF?
hverkuil: there are two separate problems here. one is hosting, the other one is the power/control structure within our community
mchehab: It would be "ours".
no, it won't be "ours"
Moving a bit south, it would seem that Estonia is welcoming towards companies and non-profit organisations of (almost) all kind, and running one doesn't seem to even require visiting the country.
mchehab: Why not, if you're a founding member?
LF is the closest thing of "ours" that we have
mchehab: no, LF isn't ours at all
I remember a not-so-fine distinction between the two was made in the Media summit...
"find a foundation in Finland"
impies to use an already existing one
s/find/found/
mchehab: I'm not proposing Finland, as the languages are limited to Finnish and Swedish.
sailus: you meand found, not find, right ?
s/meand/meant/
mchehab: Unless, that is, you're interested in learning a new language. :-)
found as in create
(10:58:29) sailus: I first thought we could find a foundation in Finland (we have two developers here), but the downside is that the languages are limited to Finnish and Swedish.
mchehab: s/find/found/
the first point would be: where such foundation would be located
sailus: and I assume that "foundation" is a generic name for "non-profit organization", right ?
I suspect we would take years just discussing that :-)
then we'll need some legal infra... accountants, lawyers, ...
mchehab: Why?
mchehab: it's not as complex as you think, and there are organizations such as https://www.spi-inc.org/ that can handle all the administrative part
Creating a foundation when all we need is someone to host a VM is total overkill.
(that's how X.org operates, they used to be a standalone foundation and had to do accounting themselves)
I'd be fine with SPI.
sorry, I got a disconnection
hverkuil: I completely disagree
based on irc logs, looks like this started an hour ago?
did the meeting time change?
We're not a political party, or have special concerns/objectives. We just need a place to store our website and git repo etc.
mkrufky: DST changed. :-o
im so confused
I suppose. Somewhere.
hverkuil: if all we need is a VM, I can give you one today, hosted in a data center, with root access
are we on the linuxtv.org discussion?
mkrufky: Yes!
mkrufky: yep :-)
pinchartl: sounds fine by me. Have 2-3 devs with admin rights there and we're done as far as I am concerned.
hverkuil: What's your opinion on who should own the linuxtv.org domain once it changes ownership?
pinchartl: using spi-inc would mean to trust them
mchehab`: that was an example, which I haven't researched personally, to show that there are options
if you don't want LF because you don't trust on external foundations, why would you be trusting SPI for that matter?
because the LF and SPI goals are very different for instance ? :-)
I'm perfectly happy if that's mchehab. I'd trust the LF as well for that.
LF is only a non-profit on paper
im trying to catch up .... but my last understanding is that at the meeting in edinburgh, we all agreed that freedesktop is a better home for us than lf
they make about $60M revenues per year
It would be nice (not sure if it is possible) if there is a second person to administrate the domain.
they have a huge financial interest in Linux
mchehab`: I'd say they have established themselves with running administrative stuff for open source projects.
and they thus have a strong political agenda
freedesktop is fine by me as well, BTW.
which I want to remain independent of
pinchartl: what's your opinion on freedesktop.org?
also, I had agreed to run the server, if thats still needed
hverkuil: fd.o would be better than LF from my point of view
(BTW, I am not wedded at all to the name linuxtv.org. Frankly, I think it is not a good name at all anymore)
it might be time to drop the name TV indeed :-)
linux-media is more than just TV nowadays
TV doesnt mean TV anymore in the industry
but than, video4linux is also a horrible, horrible name.
it is still the correct label
Now, TV means "internet video streaming"  silly millenials
that's precisely why I prefer to just discuss a better place to host it...
linuxtv.org is an establish brand that we should not consider dropping
We have Media controller, too, besides V4L2...
we're just diverging from the thing we do better: develop :-)
mkrufky: I think we should keep the domain name, but possibly not as the only one
sure
Does anyone have any objections to freedesktop.org?
I do
to provide infrastructure?
I have registered linux-media.org as it was free and at risk of being taken by domain name brokers, I'm happy to give it to any foundation we decide to register :-)
I'm not convinced why it would be any better than LF
mchehab: because they don't have the same financial and political agenda as LF
mauro, you were convinced in person in Edinburgh
the LF isn't about community anymore
it's a company trying to make money
^^^^^^^^^^^^^^^^^^^^^^^^^
hverkuil: I have no objection against fd.o. I don't know what they could provide though, that would need to be double-checked. git trees for sure, a VM with root access, we have to ask. it's worth asking though
I don't think freedesktop is worse than LF. And if there are objections to using LF (even if I don't share those objections myself), then I think freedesktop is a decent alternative.
pinchartl: I agree, we need to know what they can offer us.
We also don't need to host the git tree and the rest (mailing list and wiki) in the same location.
the thing is that trust should be earnt
from my point of view the two best options are self-hosting and fd.o. if we want to self-host, I can have a VM with root access ready today, in a reputable data center
Currently they're one and the same, but I think it'd be good to put the git trees elsewhere, if for nothing else, then security reasons.
I don't have an strong objection about freedesktop, but I don't have any reason why I should trust them more than on LF
who are the freedesktop board?
mchehab: would you prefer self-hosting then ?
https://www.freedesktop.org/wiki/
hverkuil: there are no people's names there
who is your provider, @pinchartl ?
mkrufky: I've used gandi so far, I'm fine with other options too. it doesn't matter much
before trusting on any org, we need to be confident about the ones running it
i believe that  OVH-SSD-1 is enought for us, but I would go for the ssd-2 to be sure
but i think we're better off with fd.o
mchehab: https://www.x.org/wiki/BoardOfDirectors/
Freedesktop.org comes with contributor covenants and stuff.
I'd like to read that and perhaps know a little more first.
sailus: true, with all that "maintainers should take any penalties for violations" crap
freedesktop.org is joining the X.Org foundation
Just creating a virtual machine by e.g. pinchartl is a very light-weight solution.
If we like, we can later assign the ownership of the virtual machine to a non-profit we'd found.
I don't see that as an issue really.
mkrufky: I've had bad experience with OVH in the past, they might have improved since then though
sailus, but this lighweigh solution (I actually just created one now myself) costs money ... even if its little.  it costs
mkrufky: It doesn't seem like it'd be much.
mkrufky: I'm happy to pay for the first year at least
im a big OVH customer ... im having good experience
Collabora has offered to pay as well
pinchartl: self-hosting can work
and I think it would be easy to find other sponsors too
pinchartl: that's the problem... we shouldn't have something that just one company would pay
on OVH, it costs $4-14 / month
I could probably get a budget from Samsung to pay for it, yearly based
its not a lot, but currency is not something that we have dealth with before as a video4linux team
but then Samsung would be "owning" it
*dealt
yeah thats a problem
no samsung owning it
i would rather see mkrufky own it
mchehab: someone has to pay for the costs. my ideal solution would be having a non-profit organization that can then get money from sponsors and pay for the hosting itself
or pinchartl
pinchartl: My suggestion to that is to create a non-profit in Estonia.
i think thats the right path...  we need our own entity.   THAT's work .. but its the right answer
mine is to create it in Brazil :-p\
I'm happy to sign a covenant to transfer all rights to any non-profit of our choice as soon as we have one
They take some European ID cards, and non-Europeans can get e-citizenship, both of which are good to run a non-profit.
let's be serious for one minute
if we decide to create a non-profit organization
there are two criteria
it should be hosted in a state that allows board members from anywhere in the world
Whichever coutry we pick, it must be possible for anyone of us to reasonably be a member and run the non-profit.
(although there could be restrictions based on international bans for countries such as north korea, but that's a detail)
*and*
it should be hosted in a state that allows administrative tasks to be conducted in English
we don't want to leave anyone out
neither because of where they live
pinchartl: Estonia fills both of your requirements.
nor because of what language they speak
sailus: I don't think we'll pick a location today
but I believe we should seriously think about it
pinchartl: Not necessarily. But just to say one exists.
as it's in my opinion the way forward
There could be others.
this would take a bit of time
in the meantime
I'm happy, as I've said, to pay for a VM now, with a covenant to transfer it to our non-profit organization as soon as it exists
I don't want to own anything linux-media related in the long term
too much hasle :-)
pinchartl: as I said, the payment, since the beginning, should be done by a group of different people/companies
but if it can help moving forward with hosting in the short term, I can take the additional admin load
to avoid later complaints about who owns it
mchehab: that's what I just said :-)
we should have a non-profit organization we control
If someone else is going to do the work to set up and administer such a non-profit, and finding sponsors to pay for it and the hosting, then I'm happy to go along with it. But it seems total overkill to me. I think freedesktop.org can provide the sponsoring and hosting, and they do that already for X and dri/drm.
for most companies, you can't simply donate money to someone
other companies would then sponsor that organization
and we would thus pay the bills ourselves, independently
it has to be a company/foundation to receive the donations
and pay for the VM
hverkuil: I really don't want to stick with Freedesktop CoC
as it may end by forcing use to use github/gitlab
mchehab: there's no disagreement there. what we're discussing is the schedule. if there's a need to move forward with hosting now before having a non-profit organization operational, I can help financially in the interim period with a covenant to transfer ownership to the non-profit as soon as it exists
(as that sounds to be the direction they're taking)
mchehab: it's the same as the LF CoC that the kernel adopted...
wait dont get a VM just yet
so that argument would hold against the LF too
there is technical stuff to consider about the VM / VPS that are out of scope right now (i think) ... we can talk about it but i dont want to cause more distraction.   just lets talk before you buy the VM
pinchartl: the clause that was removed from kernel CoC, plus the "helper" document made it a way better
mkrufky: pinchartl: can you guys just figure out what hosting would cost? Let's start with that.
mkrufky: I won't buy anything before we reach an agreement
and explicitly excluded external sites like linuxtv.org
lets call is "less that $20 / mon" pending some tech discussion we will have .... it can be less than $5 / mo but probably ~10
mchehab: my point is that the LF is enforcing code of conducts similar to fd.o. I think they're even worse, as they do it because of their financial interest
mchehab: I wonder if LF hosting could come with some strings such as CoC attached.
hverkuil: for that I need a ballpark estimate of the resources we need
Let's just see how expensive it is to move to a new host and get ownership of linuxtv.org.
(CPU, disk space, RAM, network traffic)
get ownership, is it not just talking to Johannes ?
sailus: the helper document mentions "kernel.org", as far as I remember, for the scope of the CoC
(for non-git commits)
the technical side is usually easier than the rest, so let's try to sort it out
call the hosting 20 / mo and its more than enough ... no worries about ram / cpu / traffic
for git commits, it is not that clear
mchehab: what's the current needs in terms of disk space, RAM and bandwidth ?
mkrufky: let's be professional about that
of course, i am just making estimations
All projects I could find have this:
https://github.com/cncf/foundation/blob/master/code-of-conduct.md
pinchartl:ram, CPU and disk space is easy to get
not sure how to collect bandwidth  data
It wouldn't be exactly the same though, if they did just provide hosting though, but I wouldn't be too sure.
directly at the VM
mchehab: can we get them then ? :-)
If the total costs are reasonable, then it is very likely Cisco would be willing to pay for it. Or perhaps 50/50 with Samsung or others.
hverkuil: would cisco give money to a non-profit organization for that purpose ?
yeah, I'm pretty sure Samsung could pay for it, provided that the VM hosting company would provide an invoice
https://www.ovh.com/world/vps/vps-ssd.xml
the smallest $3.35 /mo plan is fine ... but i would go with the middle one
$6.87 / mo
mchehab: as you correctly explained, we should avoid direct payments in the long term. it should go through a structure we own
and i would host it in Germany
if anybody worries its not enough, the #13.87 plan is double the ram and cpu
to be frank, I would prefer to have it hosted outside EU
mkrufky: Germany uses coal to produce electricity. So it wouldn't be my first choice.
due to the "right to be forgiven" crap
I'd need an invoice as well. But the total amount is (let's say) $200 a year or something, then that's peanuts and should be easy to expense for me.
ive done lots of pin testing from various locations and i am very impressed with this german datacenter
They're still building new coal power plants --- in 2018.
*pping testing
mchehab: that doesn't matter, those rights apply worldwide, but they're not as extensive as you may think. didn't you receive an answer on the ksummit-discuss mailing list about that ?
for the record, I *think* it's not a problem. I'd have to check once we have final numbers and are actually planning to go in that direction.
Another German hosting provider, Hetzner, OTOH, has a data centre in Finland. :-)
pinchartl: yes
small
I think there are plenty of good options out there.
there are plenty of VPS options, that's not the issue
pinchartl: it won't apply, for example, if it gets hosted in Brazil
the first issue is deciding on what we want to do
I mean, people in EU could file a complain, but that won't sustain under local legislation
mchehab: it will if the websites are accessible from Europe
brb
at most, they could ask their ISP to block it
I don't think we would want that, would we ?
but they can't sue the site
haven't you noticed how most websites now have GDPR policies in place, even when they're hosted outside the EU ?
pinchartl: I can't remember a single brazilian site with GDPR policies in place
not even the biggest one
mauro, its true
s/one/ones/
maybe brazil isnt the best example
I'm pretty sure larger US sites have it
:-D
we have it at vimeo
GDPR applies worldwide, as long as the website hosts private data of EU citizens
but again that's not the issue
pinchartl: EU thinks that GDPR applies wordwide
not really true
from legal PoV, one has to fill a lawsuit
mchehab: you're not a lawyer, neither am I, let's leave it there
with has to be some local law support
no more discussions about that topic please, we've already lost enough time
if a US company wants to do business with europeans, we must comply with their laws
mkrufky: very true, but linuxtv.org doesn't make business
oh, oops
the point is: if someone fills a compliant, we need to act according to local laws
(whatever it is hosted)
mchehab: do you really want to discuss legal implications of worldwide business conducted by non-profit organizations and how they're governed by international treatis ?
or to have some lawyer to take care of it
BTW: regarding making a foundation: I am not even sure I could be part of it, I would have to check with Cisco. They are paying me to do this work, so being part of a foundation might be seen as a conflict of interest. I'd definitely would have to ask permission for it.
s/treatis/treaties/
pinchartl: no, but, at the moment we go "legal" , by creating a foundation, we'll need to take those concerns
mchehab: then we'll discuss that with people who are knowledgeable about those issues
as in people who have created FOSS-related foundations
there's plenty of them
pinchartl: I'm pretty sure that, by getting a non-profit org, it will need to have some sort of legal support
mchehab: There are fewer risks there than acting as an individual.
the main point is that all of these will take a considerable amount of time...
that will diverge us from our main duties
I'm not sure why lawyers would need to be involved.
mchehab: Sakari has a point there. whoever operates linuxtv.org takes legal risks, regardless of whether it's a NPO or an individual
since the start, as someone need to prepare the foundation paperwork
all foundations need some legal document defining what the foundation will do, how it will collect money, ...
i must step away for 10 minutes, brb
mchehab: don't worry, nobody will force you to do that work :-)
sailus: yes, having a foundation will provide some sort of legal protection...
for the ones that are not part of the foundation's board
To take a step back: what is wrong with the current situation, other than that the hosting situation is rather obscure and that I don't think we own the domain?
hverkuil: thank you for getting the discussion back on track :-)
hverkuil: nothing really wrong
if I may summarize my understanding, the current issue is that we will need to move to a different host at some point in the possibly not too distant future. is that correct ?
I think talk about foundations might be something interesting for the future, but I don't see that going anywhere soon.
it has been working like that since 2005
(even before that...)
pinchartl: no, we don't "need"  to move
mchehab: I thought you were getting hints from Johannes that we should think about moving?
so why are we having this discussion ? :-)
I understood the hosting matter should be addressed soon, the administrative part can wait I suppose.
that's what I remember from the media summit...
hverkuil: there are hints that he's having less time to do any work on it
Me, too...
Sounds like a reason to think about the hosting, eh?
but he's not really pushing hard for us to move
So what should we do? Wait until the virtual machine is gone? :-)
we could keep it there for a while
I understand that it's a PC he maintains at that university?
no, it is not on a PC
Or is it a VM hosted on the university.
as far as I understang, it is on some servers inside an university
I have to admit I never understood the exact arrangement.
and he is not the one providing us the infra
the infra is maintained by another person that used to be a DVB contributor
and works at the university
hverkuil: I don't have much details about the exact arrangement
the point is: if we want to add havier load stuff there (like some build automation), we may have troubles
right now, the main demand we have is bandwidth
mkrufky probably remember that, in the past, we had some issues with bandwidth, before the infra got moved to the univestity
mchehab: as you have root access, could you start monitoring the bandwidth consumption, in order to prepare for a future move ?
the information will be useful
Perhaps we need to first get a good idea of the current setup? And who knows, perhaps the university would be willing to just host us, but on better hardware?
mchehab: It sounds like to me that the current solution may not serve as a long-term hosting solution for us.
on that time, the old provider had a max bandwidth limit
with used to be exceeded a few days before the end of the month
mchehab: Or just check host much traffic has traversed through the nic and how long it's been up.
mchehab: could you also tell what CPU, RAM and disk resources the VM currently has ? I assume that's very easy to get
I guess that would give us good enough understanding.
sailus: it would be nice to get a bit more details, but that would be a start
sailus: that probably won help much
as we may need to dimention the peak traffic
Iface   MTU Met   RX-OK RX-ERR RX-DRP RX-OVR    TX-OK TX-ERR TX-DRP TX-OVR Flg
eth0       1500 0  312432910      0 1239096 0      310109271      0      0      0 BMRU
lo        65536 0  190473719      0      0 0      190473719      0      0      0 LRU
mchehab: It'd be better than nothing. More statistics would of course be better.
 14:01:35 up 1095 days,  2:16,  2 users,  load average: 0.38, 0.32, 0.32
mchehab: How often are security updates done again?
I suspect that this is wrong
just bogus random numbers
X-)
I'm pretty sure we replaced the Debian version of the VM
maybe one or two years ago
and I'm pretty sure the VM was rebooted a few times this year
Filesystem                    1K-blocks     Used Available Use% Mounted on
/dev/mapper/linuxtv1--vg-root  49132352 43753252   2860268  94% /
again, not sure we can trust on it
~42GB of used disk space
but the number seems about right
94% is a bit worrying by the way
model name	: Intel(R) Xeon(R) CPU E7- 2850  @ 2.00GHz
4 cpus
we used to have just 2 CPUs there on another hardware
and worked well too
re: current setup, mchehab and I have root, we know ehere the data is
memory?
             total       used       free     shared    buffers     cached
Mem:       4048012    3763848     284164      52076      97036    2619292
-/+ buffers/cache:    1047520    3000492
Swap:      2093052     309356    1783696
hardware is less relevant
this machine comes from an older time
nowadays we dockerize
and use VPS
today, linuxtv.org is a vm on a box
and we have services installed directly, running directly on it
mchehab: Which kernel is it running currently?
mkrufky@www:~$ uname -r
3.13.0-43-generic
sailus: let's avoid getting into too much details
this channel is publicly mirrored
mkrufky: Wow.
LTS distro
always have old kernels
is there anything against my involvement on the new server if there should become one?
Hmm. I think there are good arguments to look for a newer hosting option. And a more active admin.
this is what i do now.
mkrufky: not from me. As long as I don't have to do it :-)
mkrufky: not from me
assuming all is ok with that,  then seriously ... nobody should think or worry too much about kernel versions or hardware specs
but we need first to get there :-)
the biggest issue is traffic
Debian oldstable has 3.16.something, and the last update was a security fix.
yes, first get there
mkrufky: I would certainly not be against not having to do sysadmin :-) we should probably have an admin in europe too though, in order to cover more time
I suspect it's from an earlier Debian release.
(and one in an asian time zone would be nice as well)
pinchartl: good point. Thank you for volunteering! :)
we can work together on it pinchartl
I said *not* be against *not* doing sysadmin ;-)
i just want to be sure its done in the most efficient way using todays cloud tech (which i know you all stay away from :-P )
oh lol
I don't remember any linuxtv downtime recently
so i volunteer to do this in my usual daily activities
(recently - over the last 3-4 years)
but i think its a good idea for someone to be familiar with the way i set it up
i should not be the only one
but again, maybe getting ahead of ourselves now
I would keep Johannes there, if he's willing to help
he has proofed his value of doing a great job maintaining it
mchehab: do you think you can ask Johannes if he can get some traffic stats?
yeah, sure
that appears the main piece of information we need when it comes to pricing.
mkrufky: I would prefer to keep the same setup, if possible
mkrufky: regardless of who sets it up, the system architecture should be a carefully reviewed decision among the admins team
hardware-wise it seems we need very little, really.
it has been proofed to work
hverkuil: yes, for the stuff we have, we don't need much
I need to leave now but I'll later read the log.
if we add KernelCI/Jenkins, then we'll need more
a lot more, depending on how we change our commit proccess
Have a nice evening (or morning or other time of day)! :)
sailus: have a nice day
I'll have to leave soon too, could we come to a closure ?
same setup is the old way
it should be containerized
im not volunteering to do old antiquated things
I would love to have a Jenkis instance that would process every submitted patch and provide some auto feedback about new warnings, checkpatch issues, compilation breakages, etc
mkrufky: I don't buy all those containerized crap
containers usually require process to run as root
.,... kernelci / jenkins  this would have to run on a different box
and a container may have older libraries
mchehab: not when they're setup correctly
with is a security nightmare
heh.   do some research
this is not what im here to argue about
mkrufky: I know all the arguments in favor of containers :-)
I'll have to leave soon too, could we come to a closure ?
we should discuss the tech aspects another day
dfor now, what did we decide?
mkrufky: agreed :-)
pinchartl: basically we need to get traffic info first since that's key to getting pricing (whatever we choose).
in practice, if I have 2 versions of the same library, packaged into different containers, I have to upgrate twice the library if it has a security flaw
sounds like we WANT to get our own hosting, but we aren't yet 100% on it
mchehab: another day
mchehab: also, this is why you wont be running the box
I guess the conclusion is that we should work on improving it
1. we want to move away from the current hosting infrastructure eventually, to add more services
but we don't have yet a clear way about how to do it
2. to do so we need to evaluate the options, and we need more info ono traffic
so, action point: get more info on traffic
mchehab: mkrufky: you're the current admins, could you work on that ?
getting traffic?
information about network traffic
i dont have access to the hosting
in only have root access to the box
how much ingress and outgress traffic we currently have
monthly averages and peaks
none of us have access to the host infra
i dont have that info
only one with access to it can get reliable data
mchehab will have to ask Johannes.
you have root access to the VM, what else do you need ?
he that owns (or rents) the physical host, has that info
pinchartl: VM hides it
it shows bogus traffic (and uptime)
what ?
no it doesn't show bogus traffic
there's a virtualized network interface
all traffic goes through that
that's all we need to konw
s/konw/know/
I suspect it shows the traffic of the entire machine
(with seems to be a very big one, as it uses Xeon E7 CPUs)
pinchartl: I guess it uses XEN
no, the virtualized nic will not show the traffic of the entire host
this VM is not up for 1095 days
that's for sure
i have another meeting in five minutes.    I think maybe it would be a good idea if Laurent and I spoke offline before next week and we can sync on what we know
I don't think XEN virtualizes the NIC
sounds good to me
eth0       1500 0  312480811      0 1239374 0      310149410      0      0      0 BMRU
the number of RX packages (312480811) is too high
I guess last reboot was last month
that would give about 120 packages/second (assuming 30 days)
I'll check that with Mike
Mike?
mkrufky:
I'm not sure if mkrufky knows about Xen
(tech details, I mean)
we'll figure it out, don't worry :-)
yeah, on Xen interfaces are paravirtualized too
https://wiki.xenproject.org/wiki/Xen_Networking
my apologies,  vimeo engineering all hands, i must go
(that's the default)
I need to go. I propose adjourning this meeting for today. does anyone second ?
(Sakari and Mike have left already, I suspect that there will be nobody left soon :-))
yeah, let's finish today's meeting
bye!
have a nice rest of the day :-)