Sooooo, I’d like to ask everybody to help and discuss the conference organization!
I’d like to follow a ‘familiar’ pattern here, building on the events I have organized before. That is, starting Friday with hacking, Saturday/Sunday talks (only lightning talks) and workshops (longer of course) and the rest of the week coding again.
Friday, we will do a meeting also with big users like universities to discuss federation and scalability, obviously anyone is welcome to join that, too, and no suits needed I will probably do a blog about it…
Our todo is still significant, probably starting with creating a nice logo
Stay tuned, I’ll start dumping my todo list here, but for now I have some other cool announcements to focus on this week. Oh yeah, this wasn’t the biggest news, not by a long shot…
I’m interested in coming, but we’re mere sysops and not devs. I know we are welcome anyway, but I would like to contribute some on the sysop Level. Here are some subjects I would like to talk about;
We’ve got a large customer, one with ~6500 users, who is going to use NC. I have a friend who works at an organistion who has 30.000. We are in a desinging phase, they work with OC and running production. They have some interesting issues, like the scalability of S3 storage, the MySQL high availability, docker support (OC didn’t support docker in the enterprise version) and some more. I would like to talk about in some more depth than on the documentation page on the design of large implementations of NC. To give something back we could update the documentation with the findings.
Different kinds of support contracts, we don’t need the fancy support contracts that are offered, we would need some big-emergency-everything-broke support contract only. But maybe that’s a question for sales and not the conference.
Different ways to support NC, we are already doing the obvious like helping out in the forums (my colleague is at least), reporting bugs, etc.
Publishing a free (as in free beer) NC service, hosted by us, delivering an alternative to all the others by guaranteeing privacy.
As you stated, this kind of subjects would be on Friday, can you confirm this? How would it look?
And let me know if i can help with the conference, we are a small company but we do have some resources.
I’m just wondering if we want to have an icon for the Nextcloud conference itself, as opposed to just reusing the app icon for the conference logo? I’m talking about something that would be inspired by the app icon, and incorporating it, but representing the goals off the conference itself (eg. presumably connecting disparate parts of the community, growing influence etc).
We’d like to do a ‘big installations & federation’ day on Friday, which will be interesting for (to be) big customers & partners of Nextcloud GmbH as well, so that sounds like worth a visit for you folks.
With regards to help, you seem to know about and have interesting scalability issues, perhaps you’d be interested in actually running a session on that subject? Of course, we’d work together on that, but it would make sure the session is more relevant for real life if it is set up by somebody who actually bumps in these problems rather than by the engineers who might think they have solved all of them already
About the documentation of scalability you offer, that is a very interesting point and it’d be very cool to do something around that together. I’ll shoot you a private message on that one.
Last but not least, we can and should discuss how one can help and get involved, that will certainly be a subject for the conference.
I would love to present a session, but the problem is this;
we are in a design phase, so no practical experience on big deployments. I read a lot and all solutions sound good on paper, but running them in production is a whole dicferent story. Theres no better testenv than production ey?
the info of the 30.000 implementation is third hand, quite usefull but i lack details. I will try to get the engineer to come with or get at least the details, but its a bit hard as he’s not that interested.
Things i know though;
oc uses the s3 storage with one bucket, so this slows down if you have millions of objects in one bucket. This could be optimized i was told. Im no fan of shared storage so i would like a distributed filesystem.
the oc ha documentation relies on one master and standby. They use galera for multimaster replication, i think this is a better design, but im curious if there are any gotchas with this setup.
we would need storage for our files, i selected glusterfs as a candidate, but i heard that this is quite unstable (node fallout and rebuild isnt as advertised) in the 30000 user env (missing details here too). Im thinking of using openstack now, but thats only a feeling and i lack experience here too.
docker used to be an issue as oc didnt support this in production. Unbelievable, but thats whats NC is for right? But probably this hasnt been in production on large environment, so what could we expect?
Again, we are very new to the oc/nc stuff and for the small en environments im not worried, but 6500 current (and 18k projected) users is a different ballgame and i have so much questions and only soso answers.
We could also do an interactive session, let people partipate in how they woukd setup such an env, or how they already have done, what their concern/wishes are, etc. Would it be an option to ask the big oc customers befkre the session and interview them?
Our idea for the event is more around very interactive workshops around certain subjects, where the organizer is mostly a guide rather than lecturing. We have talks, for sure, but those are 5 minute lightning talks exclusively, this to make sure they are on-topic and don’t bore anyone so we can all be in one room and be inspired together!
I hope to have news on a keynote speaker too and I already have a few ppl who want to give talks so I’m working full steam on the registration & CfP
including some recommended hotels. Next week or so I will reserve rooms in one for employees plus a dozen people we will probably end up sponsoring (yes, we’ll sponsor students and others who can’t afford the event on their own so easily).
Rooms:
1 big room (H1058) Friday-Monday for keynotes and talks and perhaps some of the workshops (?) ~250 ppl
Ok, thans different from what i expected, but less of a problem i think. I can talk for 5 min no problem…
So one subject could be theoretical design for large implemtentations based on what i know right now, and let people interact with that. Is that any good?
So that would be awesome for a workshop/discussion on Friday. Would rock if you can lead that.
A lightning talk on that subject for Saturday or Sunday (that’s when we do the talks, keynotes etc) would complement it nicely you can share the outcome of the discussion/workshop a bit with the entire audience.
Or I? @tobiasKaminsky what comes to my mind are two things for lightning talks. Our process with lgtm etc, gitflow… And the other one about what features had been done and what is soon to come.
Both would give people insights on how we work and what we are working on so both talks would also function as a motivator for people to join. What do you think?
I will probably do a blog about it…
