ScrewDriver1337 has quit [Ping timeout: 480 seconds]
rsalvaterra_ has joined #openwrt-devel
rsalvaterra_ is now known as rsalvaterra
goliath has joined #openwrt-devel
rua has quit [Remote host closed the connection]
rua has joined #openwrt-devel
n3ph has joined #openwrt-devel
Daanct12 has quit [Quit: WeeChat 4.4.3]
robimarko has joined #openwrt-devel
rua has quit [Quit: Leaving.]
<schmars[m]>
in case anyone else is having git clone problems recently, lower your MTU. dont know what the issue is, but it appeared in three different locations with both git.openwrt.org and github.com
decke has quit [Remote host closed the connection]
decke has joined #openwrt-devel
Shados has joined #openwrt-devel
Shados has quit []
<KanjiMonster>
schmars[m]: what does lower mean, and what is the mtu value that causes problems? without any numbers it isn't very actionable
<schmars[m]>
the standard 1500 used to work fine, then the problem appeared and i tried 1404 out of the blue, i didn't try more, have other stuff to debug :)
<schmars[m]>
other suggestions for this type of git failure didnt help (git config --global http.version HTTP/1.1, git config --global core.compression 0, git config --global http.postBuffer 1048576000)
<KanjiMonster>
schmars[m]: also v4 or v6?
<schmars[m]>
both actually - that's why i'm thinking the MTU itself isn't the issue, but lowering it might improve the circumstances so the issue doesn't trigger
<schmars[m]>
but if noone else has been seeing it, i'll book it under "freak CI issues"
<KanjiMonster>
at least pinging I had no issue with up to -s 1472 for v4 and -s 1452 for v6
<efahl>
hmm, just speculating about this... maybe github is rate- or bandwidth limiting your downloads and setting mtu smaller slows things down just enough???
<ynezz>
schmars[m]: 504 is unlikely to MTU or is it?
<schmars[m]>
sorry i linked the wrong one
<schmars[m]>
but if you look at that main_rockchip/armv8 failure 15 minutes ago, i've been having that same issue for the last few days, where the clone in scripts/feeds aborts because the connection is closed prematurely
<efahl>
yeah, gateway timeout seems very unlikely as response to rate limiter
<schmars[m]>
Cloning into './feeds/luci'... \n error: RPC failed; curl 18 transfer closed with outstanding read data remaining \n fetch-pack: unexpected disconnect while reading sideband packet
<schmars[m]>
this is what i mean, nevermind the 504
skynet2_ has joined #openwrt-devel
<schmars[m]>
happens with github.com too
<schmars[m]>
and same with git.openwrt.org
<schmars[m]>
i tested MTU=1404 as a wild guess and it disappeared. haven't tried other values, no deep debugging time today
<ynezz>
its random, so you would need to run this for quite some time
<ynezz>
usually happens when the server is overloaded
skynet2 has quit [Ping timeout: 480 seconds]
<schmars[m]>
so you've seen this in the past?
<efahl>
but if it happens on both github.com and git.owrt.org, sounds like network rather than server?
<schmars[m]>
i had never seen it before
<schmars[m]>
yeah network, or git itself on the server side
<schmars[m]>
don't know if git on git.openwrt.org has been updated, but git client on my workers definitely hasn't
<schmars[m]>
oh well it might actually, given they're building in a fresh container every time
<schmars[m]>
anyway i can help debugging and have reliable reproductions, but not today :-)