multireq / v04x / IPNS issues

[ghost]

06:50-0400 <holodisc_> anyone else experiencing errors when using the ipns gateway http://gateway.ipfs.io/? Seems like every 5th request results in error 400 "Path Resolve error: could not resolve name."
06:50-0400 -- holodisc_ is now known as holodisc
06:51-0400 <holodisc> not exactly every 5th, but approximately 
06:51-0400 <holodisc> sorry i meant http://gateway.ipfs.io/ipns
06:52-0400 <@lgierth> that happens the IPNS name can'
06:52-0400 <@lgierth> t be resolved
06:52-0400 <@lgierth> are you on 0.4.1?
06:53-0400 <holodisc> lgierth: but i'm using the same hash over and over, and it does resolve most of the time
06:53-0400 <holodisc> yeah 0.4.1 on my end
06:54-0400 <@lgierth> mh can you check which of the servers it hits? ipfs.io (and gateway.ipfs.io) are backed by 7 servers
06:55-0400 <@lgierth> i assume you might occasionally hit one which can't resolve it for some reason
06:55-0400 <holodisc> lgierth: sure let me check
06:59-0400 <holodisc> lgierth: no clear pattern... it failed for 104.236.179.241 now it works for that ip
07:00-0400 <@lgierth> ugh
07:00-0400 <@lgierth> could you try v04x.ipfs.io?
07:05-0400 <holodisc> lgierth: slooow, but eventually returned my data
07:07-0400 <holodisc> lgierth: uhm yeah, response time are sometimes fast, sometimes super slow on v04x, but at least it never returns 400s

[ghost]

07:11-0400 <@lgierth> ok i'll look into it this week, there seems to be two issues
07:11-0400 <@lgierth> one being IPNS itself, the other being multireq, our v04x/v03x multiplexer
07:12-0400 <holodisc> lgierth: cool :)
07:13-0400 <holodisc> lgierth: fyi here's my test:
07:13-0400 <holodisc> for n in $(seq 100); do curl 104.236.176.52/ipns/QmcPd9peA1aqf9NHUBvVZoYdxvM3sQkTZqXAq3YuBMxpbb; sleep 1; done | sort | uniq -c
07:13-0400 <holodisc>   94 Hello world!
07:13-0400 <holodisc>    3 Path Resolve error: could not resolve name.Hello world!
07:13-0400 <@lgierth> tracking here: https://github.com/ipfs/infrastructure/issues/160
07:13-0400 <holodisc> lgierth: this just in: 94 + 3 = 100
07:14-0400 <@lgierth> hehe funny
07:14-0400 <@lgierth> oh yeah
07:14-0400 <@lgierth> the 400 error doesn't print a newline
07:14-0400 <@lgierth> so it counts that and the next response as one
07:14-0400 <@lgierth> so there's another issue right there haha
07:15-0400 <@lgierth> minor though
07:15-0400 <holodisc> lgierth: ahhh explains it
07:15-0400 <holodisc> lgierth: lgierth didn't see the Hello World at the end of the line :D
07:16-0400 <@lgierth> yeah i got a similar ratio: 98:2