03:08 <wont> Hello

03:09 <wont> I was struggling to get jane st core working in mirage. I was hoping someone would have already done this or would be able to tell me why it's not possible.

03:10 <wont> When including core in the packages list mirage build errors out with "ocamlfind: Error from package `threads': Missing -thread or -vmthread switch

03:10 <wont> Command exited with code 2.

09:46 <hannes> wont: nobody ported `core` to MirageOS, threads are not available on MirageOS... there is no recent activity on that front, although an open issue https://github.com/mirage/mirage/issues/282

12:30 <dstolfa> hannes: Well, multithreading in kernels is always kind of difficult :P

12:30 <dstolfa> Also depends on how you want to assign and figure out what your unikernel would be running, SMP/NUMA,...

12:31 <dstolfa> How you would schedule them, it's a whole task to be perfectly honest, and not an easy one at that

12:31 <dstolfa> them being kernel threads, that is

12:32 <hannes> dstolfa: I like simple things... cooperative multitasking on a single core. I believe we can solve 99% of all problems with that (and scale using marshalling (netmap/some shared memory/whatever is not there, or simple TCP for now) and multiple virtual machines)

12:33 <dstolfa> hannes: I fully agree, multithreading is something that should be evolved over time if a need for it develops in the case of MirageOS

12:33 <dstolfa> hannes: It adds very high complexity

12:35 <dstolfa> I've recently had the privilege of dealing with a highly multithreaded environment, which also has to be handled with great care in the case of NUMA, I still have race conditions, but it's getting stable :)

12:36 <hannes> sorry to hear that, I debugged multi-threaded code for way too long in my lifetime

12:37 <dstolfa> hannes: Heh :)

12:38 <dstolfa> hannes: Well, in this case it was 4 queues, arbitrary interrupt threads, task queues, communicator thread(both guest and host)

12:38 <dstolfa> and kernel events towards userspace on the host side

12:38 <hannes> my brain is simply not smart enough to cope with that...

12:38 <hannes> ;)

12:39 <dstolfa> hannes: I started doubting this statement: if (cond && !cond) is the same as if (cond) if (!cond)

12:39 <dstolfa> well, cond1, cond2

12:39 <dstolfa> That's the effect concurrency has on people when there's too much of it :D

12:48 <kensan> hannes: We have the same line of thinking and we have not had a real usecase for SMP guests on Muen.

12:49 <dstolfa> kensan: Well, if you implement SMP, users will inevitably want NUMA in some cases as well

12:50 <dstolfa> Then some will come along with SMT, and you'll end up with a bunch of code :P

13:44 <lobo> hannes: i've seen that graylog supports rfc5424. it is free software, but has jre, mongodb and elasticsearch as dependencies

13:45 <lobo> hannes: haven't tried it myself yet. but some fellow networkers claim that it is easier to install than elastics/logstash/kibana

13:45 <lobo> https://www.graylog.org/

13:46 <hannes> lobo: thx for investigating!

13:48 <hannes> hahaha, and there's even a friend of mine (with whom I studied) who works (worked?) on graylog

13:51 <kensan> dstolfa: I am *really* not keen on adding that much complexity ;)

14:09 <dstolfa> hannes: :)

14:10 <dstolfa> kensan: oh, wrong ping, sorry hannes

14:10 <dstolfa> You guys have a similar name and I've been awake since 5

14:10 <dstolfa> (11 hours ago)