IRCaBot 2.1.0
GPLv3 © acetone, 2021-2022
#ls2
/2022/12/23
orignal back to vanity miner
orignal implemented for opencl
orignal if anybody is interested let me know
orignal our release will be major
orignal because no more SSU
dr|z3d got a new curious console bug, zzz. looks like a hard limit in jetty's form processing function. cake.i2p/view/vrKmHN7Ezz_QqkqCoyiadsxM6vzrBjYyGSsr2fo6F_yaRxtZbxZM/vrKmHN7Ezz.txt
dr|z3d possible mitigation would be to only submit changed fields, dunno?
dr|z3d or maybe it just doesn't like large forms.
zzz I only see 8 form items per tunnel, plus one hidden for 9, so that would support 111 tunnels for a limit of 1000, unless you've done some hacking
dr|z3d no hacking, just testing hosting limits.
zzz ok, so you found the limit
zzz speaking of limits, I have a favor to ask
zzz would you please change your ParticipatingThrottler limits to be as strict as those in canon?
zzz as we worked out with obscuratus
zzz so yours and stormy's routers aren't part of the problem
dr|z3d not really, I've just found the limit for the form submission :)
dr|z3d elsewise, doesn't appear to impact anything.
orignal what are you talking about?
zzz java-specific stuff probably better in #i2p-dev next time dr|z3d
zzz java stuff orignal
dr|z3d re -dev, sure. my bad.
orignal I thought the oriblem with number of tunnels
dr|z3d as for throttler limits, they're now at around 20 requests in 220s, so shouldn't be part of the problem.
zzz to -dev for this one
dr|z3d or you think that's too high? it's catching a chunk of offending routers.
orignal too small
orignal I would limit to like 2 or 5 per second
dr|z3d yeah, but are you doing any limiting, orignal? :)
orignal not for participation
orignal only by number and bandwidth
dr|z3d might be time to consider some sort of request throttle given the ongoing network issues.
orignal maybe
orignal like 500 per minute
dr|z3d 500 per minute? that's an insane amount.
zzz not really, that's 5000 tunnel requests per 10 minutes
orignal 5K per 10 mintes is fine
orignal in my opinion
dr|z3d then I'm confused. we're currently allowing somewhere in the region of 20 requests per 220 seconds, in I2P+, and similar in I2P.
dr|z3d that's per router, not total.
orignal ofc I meant total
dr|z3d I think we do some throttling of total requests if they ramp too quick, but the issue we've currently got is not very many routers on the network making a ridiculous number of requests in a short period, which appears to be impacting overall net perf.
orignal if I'm a router and produce bunch of requests
orignal they would go trough different paths
dr|z3d well, you'd hope so.
dr|z3d and if a router decides it wants you to handle a huge number of requests, then it should politely (or impolitely) told to take its business elsewhere.
orignal it's worthless to limit per router
orignal because they would come from different
dr|z3d yup, that's true. but you're still able to throttle any router that's making excessive requests, regardless of where those requests originate. which isn't a bad thing.
dr|z3d and with the throttle calibrated correctly, you won't be throttling much.
zzz peer test corner case FYI - my bug -
zzz I was alice, connected to bob for 3 hours, started a peer test with bob
zzz but my temporary IPv6 address had changed an hour before. My IPv6 address for the session with bob was now deprecated
zzz I tested my old deprecated address, not my new one, so results came back SNAT
zzz hopefully fixed
orignal you mean one in your RI?
orignal how it's possible?
orignal if you keep connecting
orignal in mean time 850 Mh on RX580
zzz deprecated addresses are still good for a week on linux
zzz so there was no path migration