All boardsБлинчикиInternationalКрендельнаяЛес

Тред разработки борды (Board development)


#123592

Тред разработки борды (Board development)

1080×10771.50Mb

Прошлый (Previous) >>4410 ➡


Формат смайлов (Smiles requirements) >>122797 ➡

Редактирование профилей (Idol profiles editing) >>123590 ➡

114 posts and 49 files omitted
See all
#467299Кагами

>>467214

>

I noticed this png of all smiles is sent on first load of a thread or board or all. But I could not find where in the files exactly this is done.

browser downloads it once it encounters smile in the thread via background-image rule in the CSS

>

that smiles.png is 644kB and thus very wasteful for mobile data

it's 2.1M for retina users btw (2x version)

>

on ever pageload where it isn't cached already

it's being requested only once, then it should be cached forever by the browser (well at least for several months)

>

would it not be better to serve mobile users the smile pictures individually

it was done intentionally that way...

there's request/metadata/parsing/etc overhead when you download small images individually

>

since most smiles are almost never used

it's possible to remove unused smiles

also maybe split smiles in groups, so you download smiles-part1.png, smiles-part2.png, etc, once you see it

right now it shouldn't be a huge problem...

>

AllSmiles.png is by far the one thing that makes site load slower

it doesn't freeze other network requests, you just don't see the smiles until the sprite is loaded, also it happens only once

#467322

>>467299

>

then it should be cached forever by the browser (well at least for several months

oh alright that makes it less of a problem, I didn't know how long it would be cached.

>

maybe split smiles in groups, so you download smiles-part1.png, smiles-part2.png, etc, once you see it

this would make sense, just putting the 50 most used smiles in one and less used one in another. You're right if it gets cached that long it wont really matter for browsing but I guess first time users would still benefit.

>>467301

thanks for showing me


Another thing:

I've just looked a bit into how the post feeds work but on mobile it is not the best experience. whenever you put browser into background and switch to another app and then at some point resume, most posts that were made in the meantime don't show, usually only the very latest post is updated. this leads to a lot of conplete reloads of the page or missed posts.

Do you know why this happens?

sorry for dumb questions btw gotta learn golang a little bit

#467407

>>467322

>

put browser into background and switch to another app and then at some point resume, most posts that were made in the meantime don't show, usually only the very latest post is updated

yeah, thread synchronization via websocket is badly written, it requests only for ~5 latest posts after reconnect

need to rewrite it properly...

>

gotta learn golang a little bit

#467466

>>467407

I wonder if this behavior still exists in meguca, would be something weird to overlook considering how many features they have...

maybe they are doing something else entirely, I can't really tell

#467577

>>467466

fork was made a long time ago, meguca has changed a lot since then, so it's impossible to merge

fix is not that hard, I'm just don't have enough time for that currently

#467582

>>467577

>

I'm just don't have enough time for that currently

don't worry about it, chingu

#471658

Это фиаско, братан

#472573

Are deleted posts actually deleted or are they saved somewhere in the db?

#472599## Admin ##

>>472573

they could be saved in backup if it was older than 1 day, otherwise actually deleted

#472637

>>472630

no but it would have been cool to have all posts including deleted ones as a dataset.

#472645

>>472630

>>472637

*forgot to add:

if you can db.getallposts() and save it in some useful way(.csv or .txt) and upload on an uploadsite I would be very grateful If not it's ok, I will just have to do a little scraper.

I want to use the posts for sentiment analysis and other stuff


#472656

>>472645

what boards do you need?

I can dump post_id,date,post_body information

#472657

>>472656

/int/ and /wood/

reply count is not available right? that would be interesting too but I can do that manually later on

#472668

>>472657

>

reply count is not available right? that would be interesting too but I can do that manually later on

it's available but you can build such map by yourself, just by searching for ">>\d+" in post body

#484698

hey kag, we finally made :smiles: changes. pls accept pull request when you have time

#485340Кагами

>>484698

sure, thanks, will do a bit later

#485344Кагами

>>485342

2 hours downtime because of docker update, service wasn't able to restart itself correctly...


Oct 18 06:57:03 kpop.events systemd[1]: Starting Daily apt upgrade and clean activities...
Oct 18 06:57:12 kpop.events systemd[1]: Stopping cutechan service with docker compose...
Oct 18 06:57:12 kpop.events systemd[1]: Stopping kpopnet service with docker compose...
Oct 18 06:57:14 kpop.events docker-compose[1759187]: Stopping kpopnet_kpopnet_1 ...
Oct 18 06:57:14 kpop.events docker-compose[1759186]: Stopping cutechan_cutechan_1 ...
Oct 18 06:57:15 kpop.events kernel: veth0037635: renamed from eth0
Oct 18 06:57:15 kpop.events systemd-networkd[693]: veth0a28b98: Lost carrier
Oct 18 06:57:15 kpop.events kernel: db: port 2(veth0a28b98) entered disabled state
Oct 18 06:57:15 kpop.events kernel: db: port 1(veth7d1226e) entered disabled state
Oct 18 06:57:15 kpop.events kernel: device veth7d1226e left promiscuous mode
Oct 18 06:57:15 kpop.events kernel: db: port 1(veth7d1226e) entered disabled state
Oct 18 06:57:15 kpop.events systemd-udevd[1759138]: vethfce82cd: Failed to get link config: No such device
Oct 18 06:57:16 kpop.events docker-compose[1734582]: cutechan_cutechan_1 exited with code 2
Oct 18 06:57:16 kpop.events docker-compose[87257]: kpopnet_kpopnet_1 exited with code 2
#493089

Добрый день, мистер Кагами. Хотелось бы уточнить, а точнее попросить у вас обновить "Борам". Это возможно?

#493267

>>493262

Ютуб длл. Или это не поможет? По опросам, многие используют, но без возможности прямой скачки через сам конвертер. Не очень удобно если честно.

#493346

>>493345

Ты меня переоцениваешь. Но я попробую.

#493519

>>493345

Когда на гитхабе обновите примерно?

#508628

since vlive added this url format it would be nice for them to also embed

for example https://www.vlive.tv/post/1-19990273

you can still get the videoSeq from embedded json, beware that this url format might not have videos: https://www.vlive.tv/post/1-19758287

seems like the embeds still work the same once you have the videoSeq >>508512 ➡

{
  "postDetail": {
    "post": {
      "url": "https://vlive.tv/post/1-19990273",
      "title": "이런저런💬",
      "postId": "1-19990273",
      "officialVideo": {
        "videoSeq": 224269,
        "type": "VOD",저런💬",
#574623

did server-nim go down last night?

do you need more bitcoins?

#574642## Admin ##

>>574623

Out of memory: Killed process 3022967 (kpopnetd)

cutechan was fine though. what kind of error did you get?

#574643

>>574642

server was timing out for about 1-2 minutes 2 times at around 02:00 cst

#574654

>>574649

you put me in a deep thought


I see, so you had slow response at 00:51 UTC, but I can see OOM killer at 01:53 UTC and nothing at 00:51

could be coincidence, just some packet drops

#574655
1920×10807.74Mb00:17

>>574654

glad it wasn't something serious then

#578096

Admin, I appreciate whatever you do that allows me to post from vpn, yet the boards isnt overrun by psychos and retarded LEOs wasting tax money trolling those psychos.

#586822## Admin ##

40min downtime because ssh updated default HostKeyAlgorithms to prefer ed25519 so sftp couldn't connect because of the different public key fingerprint

#649010## Admin ##

node was rebooted because of hardware issue but postgresql wasn't properly started for some reason so 6.5hours downtime

#699526## Admin ##

media domain was changed back to up.kpop.re

I will create permanent redirect from store.kpop.events to up.kpop.re soon and eventually dismiss kpop.events domain

please report any issues you encounter


>>691720

thanks, will add a bit later

#700018

>>699526

>

thanks, will add a bit later

we also have some in this thread >>670278 ➡

might be better to add and remove more at once at a (later) time

#738492

So what exactly is the problem with missed posts not being fetched on mobile sometimes?

I just looked at it briefly, but shouldn't prepareToSync() be triggered in /ts/connection/state.ts onWindowFocus() if the state is desynced?

#738681

>>738492

iirc backend can return up to 5 missed posts only in websocket handler

that can be fixed by adding request handler to return posts since the last available on the client and do that request after reconnect

#745760## Admin ##
534×80067.78Kb
534×80083.98Kb
534×80096.01Kb
534×80079.37Kb
534×80079.50Kb
534×80069.18Kb

6 images per post

For convenience with IVE and LE SSERAFIM. Hope UI won't suck too much.

Could be 7 too because a lot of groups are 7, but better to stop at some point to avoid too many images per post mess.

#745961

>>745760

>

For convenience with IVE and LE SSERAFIM.

and now OMG

#762531

Вопрос который уже очень долгое время интересует всех.

Когда boram обновят?

#767274

quite long downtime just now 220713 ~09:00 KST

#767291## Admin ##

>>767274

I can see connection issues to the store host from 02:35 UTC+3 till 04:21 UTC+3, i.e. it wasn't possible to upload images

have you noticed anything else?

#767365

>>767291

whole site was not responding for about 15 minutes, then again for a second time an hour later. you can see the post „back alive“. it happened before and after that

#767448## Admin ##

>>767365

Seems like network connectivity issue on the hosting side, I just experience the issue but with the main site.

They don't have any incidents reported on the status page though

I created ticket to get some info.

#767464

kagami plz fix I can't live without this site

#767467

>>767464

small deprivation will make you like it even more though

#767542
>

Your VPS with is being target of DDoS attacks. It has stopped after since 2,5 hours ago.

>

Our system automatically blackholes an IP in case the DDoS is large enough to avoid connectivity issues for other customers.

>

The blackhole is being removed once the DDoS has stopped for at least 5 minutes.

#798526## Admin ##

Someone is posting pornography in old threads, trying to abuse the site.

Added IUC and bella-fan as moderators to all boards.

#892368
64×644.64Kb
32×322.04Kb
64×644.53Kb
32×321.99Kb
64×644.49Kb
32×322.06Kb

Hi, I would request that some new smiles to be added

:eunchae_lol: :sana_wut: :winter_finger: (to have it match with )


Thank you for your consideration

#892369
64×644.39Kb
32×321.98Kb

And also this one

:isa_wut:

#892376

>>892368

love the winter finger one. :icy_touch:

not a fan of the other ones, they seem kinda blurry

#1014153## Admin ##

Upgraded docker container and dependencies of the server.

Please report if something has broken.

#1015969

>>1014153

As I said in the Wood thread, maybe you didn't see it there

The image loading has been slow for the past 3 days

Significantly so in some cases

Let me know if I can do somethin to help you find out what is wrong

#1016083## Admin ##

>>1015969

I've also noticed it, but thought it's my ISP.

Shouldn't be related, might be some issue on storage node. Will check.

#1016086## Admin ##

>>1016083

Disk performance seems to be same as initially.

Could be the network path between storage node and Cloudflare.

#1016092
% for i in `seq 10`; do time curl -so/dev/null "https://up.kpop.re/src/1a/6bd4e3454f8768863072bfe3a4982b04b75ec1.jpg?`date +%s%N`"; done
curl -so/dev/null   0.03s user 0.01s system 17% cpu 0.207 total
curl -so/dev/null   0.03s user 0.01s system 16% cpu 0.236 total
curl -so/dev/null   0.03s user 0.01s system 22% cpu 0.170 total
curl -so/dev/null   0.02s user 0.01s system 23% cpu 0.149 total
curl -so/dev/null   0.02s user 0.01s system 21% cpu 0.157 total
curl -so/dev/null   0.03s user 0.01s system 25% cpu 0.148 total
curl -so/dev/null   0.03s user 0.01s system 24% cpu 0.168 total
curl -so/dev/null   0.03s user 0.01s system 23% cpu 0.161 total
curl -so/dev/null   0.02s user 0.01s system 21% cpu 0.153 total
curl -so/dev/null   0.02s user 0.01s system 23% cpu 0.153 total

seems normal, I've also tried seq 100. mysterious issue

#1016093## Admin ##

Oh, I've got an idea!

% time quiche-client https://up.kpop.re/src/1a/6bd4e3454f8768863072bfe3a4982b04b75ec1.jpg >/dev/null
quiche-client  > /dev/null  0.03s user 0.01s system 0% cpu 5.244 total

that might be it, UDP performance!


I will disable HTTP/3 for the cloudflare CDN, will see if that helps.

#1016163

>>1015969

can you try right now


% for i in `seq 10`; do time curl -so/dev/null "https://up.kpop.re/src/93/b9f9e327da7c9dae8e4e048a25ba36f0896dad.jpg?`date +%s%N`"; done
curl -so/dev/null   0.05s user 0.01s system 23% cpu 0.260 total
curl -so/dev/null   0.04s user 0.01s system 23% cpu 0.249 total
curl -so/dev/null   0.04s user 0.01s system 23% cpu 0.236 total
curl -so/dev/null   0.03s user 0.01s system 17% cpu 0.238 total
curl -so/dev/null   0.02s user 0.01s system 18% cpu 0.182 total
curl -so/dev/null   0.04s user 0.02s system 24% cpu 0.226 total
curl -so/dev/null   0.05s user 0.02s system 2% cpu 2.492 total
curl -so/dev/null   0.05s user 0.01s system 28% cpu 0.202 total
curl -so/dev/null   0.05s user 0.02s system 4% cpu 1.518 total
curl -so/dev/null   0.05s user 0.02s system 2% cpu 2.331 total
#1016164

>>1016163

seems the issue with my local ISP this time

with mobile ISP I get stable 500-700ms

#1016186

>>1016179

>

but afterwards every following webm was loading normally

that's to be expected because of the caching

not sure why initial loading is slow...

#1016198

>>1016195

unfortunately I can't see any error in nginx log on the storage node

it's just one request in HTTP log, I guess it's all handled inside cloudflare cloud from that point (exchanging between different nodes)

might be some timeout issues but dunno how to debug that because happens randomly

#1016199

>>1016195

>>1016198

added $request_time to the log, please report if happens again, maybe it's related to when storage answers too slow


or maybe cloudflare isn't useful as CDN at all...

#1017431## Admin ##

>>1016195

Pictures/videos were loading very slow for me in the morning (but fine if requested directly).

Now it's very fast again.


My only theory is that it's issue on the cloudflare's side which I can't control unfortunately. Only by changing it to something else but dunno with what. It's mostly fine except the random performance issues.

#1019621## Admin ##

>>1017431

I noticed something interesting. In Chrome media is still loading extremely slow, but tried in another browser and it's fast there for some reason. I've started investigating and found this:


https://community.cloudflare.com/t/slow-when-on-cloudflare/544387

then I checked https://up.kpop.re/cdn-cgi/trace


For some very weird reason, I have different colo values in different browsers, running on the same computer. So it seems traffic for Chrome goes through very busy path. Will investigate further.

#1019633

>>1019626

tried restarting browser, changing language flags, copying request as curl from devtools and running in console, running chrome with --disable-http2. after some time the colo changed to a good one (very close to my geographical location) and images are loading fast. seems like that's what happened here >>1017431

it's very random, all hidden inside cloudflare's logic of selecting CDN node for the current request. and the HTTP2 makes it even worse, because chrome keeps connections open


now I'm 100% sure it's cloudflare's issue. but don't know how to fix it, especially on a free plan

#1020391

>>1019646

Doesn't seem to cause much difference. Still very slow in case of colo=SIN. And I can't even report that to Cloudflare:

>

Customers on our Free plans can only submit tickets for billing, account, and registrar issues.

Maybe better to abandon it completely because it causes more issues than helps, will think about it.

#1022208

>>1020391

>

Doesn't seem to cause much difference.

Actually it seems a bit better that way. (Unless it's random coincidence.)

If caching is disabled and colo is bad, then all images are loading, just slowly.

If caching is enabled and colo is bad, then it may stuck for many seconds.

So I will keep it disabled for now.

#1028825## Admin ##

Storage configuration was updated (removed cloudflare)

Please report errors with media files (if any)

#1084379

>>1028825

Pics and videos have been loading somewhat slow for me past few days

Thought it might be just me but it happens on 3 different connections at home, at work and on mobile

Could you have a look if something might be happening?

#1096237

Кагами почисти гуро спам

#1102908

kagami is there a convenient solution for banning phoneposters available?

sadly annoying people tend to be phoneposters and used to ban evading already

#1103262## Admin ##

>>1102908

you can send me email if there's a problem

sadly the current mod interface allows only simple single IP bans which of course not enough in such cases