On Tue, January 30, 2024 3:25 pm, Bruce Jagid wrote: >> I'm also not a OS dev >> cannot the OS do some testing/benchmarking >to get a grasp on what the > limit >> could be? YOU are the OS in your example, and you >would know the limit when >> you > would do >> curls slower and maybe you would get more >and more pain.. and crash in your >> example would be your >muscle being in such pain you > wouldn't >> be able to do anything with your >arm/whatever > > So your body automatically benchmarks how many bicep curls you can do in an > hour without you having to think about it? You use your body to measure the > bicep curls it can do, it doesn’t automatically do that. You can use your OS > to perform the benchmark, but to expect the OS to designate resources > automatically to benchmark itself is equal portions naïve and obtuse. You have > a very specific use-case, you should do the work to find your answer.
it can know limit more-less, yes, based on earlier curls maybe not automatically, but having a utility that does this for you and you can run it once after each hardare change to find out, but I am not sure you say it depends on use-case, I do not understand what you mean if you read my earlier replies, you would find out that I said I already tried searching online for like 1 hour, there is some sort of crazy formula one dude did a lot of math, snipets from code, is that what you mean? because what you say sound like there are multiple types of FDs, maybe network FDs and normal FDs? - best regards > > > On Tue, Jan 30, 2024 at 10:20 AM <beecdadd...@danwin1210.de> wrote: > > >> I'm also not a OS dev >> cannot the OS do some testing/benchmarking to get a grasp on what the limit >> could be? YOU are the OS in your example, and you would know the limit when >> you would do curls slower and maybe you would get more and more pain.. and >> crash in your example would be your muscle being in such pain you wouldn't be >> able to do anything with your arm/whatever >> >> so you're saying the only fucking way to know a true hardware limit is the >> worst that could be - a crash??? what if crash doesn't happen right away? in >> my case hardware ISP router could be limiting the potential of i2pd software >> or torrenting software boom corrupted data, processes, uncompleted important >> work, lost important work, pain in ass, etc literally couldn't that corrupt >> the entire system, a crash? >> >> tell me I am worrying too much, but even then a crash is the worst thing >> someone can rely on, I think it's unprofessional that the OS allows for that >> sort of insecurity if all you said and I said is correct, I consider that >> to be a security vulnerability at least, not to mention other >> vulnerabilities >> >> On Tue, January 30, 2024 1:32 pm, Bruce Jagid wrote: >> >>>>>> >> >>>>>> like I asked and no one answered: where >>>can I check HARD LIMIT >>>>>> of >> my >>>>>> computer? >>>>> >>>>> you can't really. you can try increasing >>until you run into >>>>> problems >>> and back >>>>> off a bit, but it probably depends on what >>else the kernel is >>>>> doing. >>> usual >>>>> approach is to restrict the software to >>using the resources that >>>>> you >>> expect it >>>>> to actually need and restrict it from making >>more demands than that >>>>> >> to >>> orotect >>>>> the rest of the system. >>> >>>> this sounds like a bug to me hard limit must be known, else is like >> playing >>>>> cards, you never know when >>> you >>>> lose (you crash) and no one answered my question yet about >i2pd's >>>> connections to other >>> routhers >>>> with can well surpass 8192 up to +30000 >connections, and if I am right >>>> >>> then >>>> each connection needs a FD? I worked with >networking and programming a >>>> >>> little, >>>> so this makes sense to me can anyone >verify? if yes, then yes this is >> a bug >>>> and I am >disappointed that the only way is >>> to >>>> run blindly and trust before crash >>> >>> I might be out of line here since I’m new to OS dev stuff, but what >>> >> you’re >>> asking doesn’t really make sense to me. A file descriptor is a software >>> abstraction built onto the hardware and the exact implementation changes >> from >>> case to case dependent on hardware. It’s like if I asked my doctor “give >> me >>> the exact limit of bicep curls I can do in an hour.” In the same way the >> body >>> has no conception of a bicep curl(only the fatigue from moving), the >> hardware >>> doesn’t know what you mean by a file descriptor(only the residual >> resources >>> needed to maintain one), and there’s like 20 ways of doing a bicep curl, >> so >>> demanding such a concrete hard limit number makes no sense. >>> >>> - Bruce >>> >>> >>> >>> On Tue, Jan 30, 2024 at 6:52 AM <beecdadd...@danwin1210.de> wrote: >>> >>> >>> >>>> On Tue, January 30, 2024 11:23 am, Stuart Henderson wrote: >>>> >>>> >>>>> On 2024/01/30 10:53, beecdadd...@danwin1210.de wrote: >>>>> >>>>> >>>>> >>>>>> I see the confusion I made I am sorry, when I said routers crash I >>>>>> meant actual ISP hardware routers. >>>>> >>>>> For an ISP "customer premises equipment" router (home/officr router)? >>>>> That often means you made too many connections and exceeded the size >>>>> of NAT/firewall state table that they can cope with. Also for ISPs >>>>> with CGN, you might have a limited port-range that you're allowed to >>>>> use and can't make more connections once that has been exceeded. >>>> >>>> is there way to verify it's the 1st thing, which can be fixed by custom >>>> router, yes? any computer with 2 NICs can be a OpenBSD router, yes? I >> seen >>>> people do that, is cool >>>> >>>>> >>>>>> like I asked and no one answered: where can I check HARD LIMIT of >>>>>> my computer? >>>>> >>>>> you can't really. you can try increasing until you run into problems >> and >>>> back >>>>> off a bit, but it probably depends on what else the kernel is doing. >>>> usual >>>>> approach is to restrict the software to using the resources that you >>>> expect it >>>>> to actually need and restrict it from making more demands than that >>>>> to >>>> orotect >>>>> the rest of the system. >>>> >>>> this sounds like a bug to me hard limit must be known, else is like >> playing >>>> cards, you never know when you lose (you crash) and no one answered my >>>> question yet about i2pd's connections to other routhers with can well >> surpass >>>> 8192 up to +30000 connections, and if I am right then >>>> each connection needs a FD? I worked with networking and programming a >>>> little, so this makes sense to me can anyone verify? if yes, then yes >> this is >>>> a bug and I am disappointed that the only way is to run blindly and >> trust >>>> before crash >>>> >>>>> >>>>>> what it depends on, on CPU? where is utility that shows max FDs, >>>>>> and per-running-process FD usage and their max setting? if this does >>>>>> not >>>> exist, >>>>>> I think why not? >>>>>> I think if user has to manually set FD limits and know potential of >>>>>> >>>>>> >>>> programs >>>>>> and OpenBSD and hardware, where is utility to help with that? I did >>>>>> >>>> search >>>>>> on the internet, all shit.. >>>>> >>>>> fstat shows per-process FD use, but the kernel backend for it is a >>>>> bit >>>> buggy >>>>> and can sometimes crash the kernel, so it is best to avoid running it >>>>> >> on >>>> an >>>>> important system. >>>>> >>>>> >>>> >>>> oh really I probably cannot verify the usage of I2Pd if it exceeds 8192 >>>> because my router goes stupid and crashes, can you? if you can't I'll >> give it >>>> a try, please tell me if you can.. I would try increasing bandwidth >> speed to >>>> X and transit tunnels to maybe 10k, try with >>>> afloodfill maybe, too.. because even many tunnels - there can be many >> to 1 >>>> i2pd peer(i2pd router) which translates to 1 FD, right? and if you go >> to web >>>> console of i2pd and go to Transit Tunnels tab, you can see => [some >> number >>>> like ID] 5.0 KiB, and then you see more of same, but the arrow '=>' is >> not >>>> there, so that maybe indicates it's the same peer/i2pd router that the >>>> following tunnels are to/from.. most have 1 tunnel, some have 6 >> tunnels, a >>>> lot have 2 tunnels >>>> >>>> but I am not getting FD count with fstat, the number is not the same >> with >>>> 'Routers' in web console of i2pd, so maybe I was wrong >>>> or maybe i2pd recycles FDs to be much better at efficiency so it has >> Routers >> >>>> stored addresses somewhere, and makes connections only if needed (which >>>> >> take >>>> up FD spots) >>>> >>>> >>>> >>>> >>>> - best regards, I like talking to you, you care about this and want to >>>> help, it can be seen >>>> >>>> >>> >> >> >> >
