report saying network is unreachable + connection refused

[agenteo]

Hi,

first of all great tool I am loving it.

I am running a rate 20 attack on a heroku app (mine) on a test plan of around 2K urls for about 1H.

when the reports completes I am seeing lots of:

 dial tcp: read udp 172.25.64.30:53: connection refused
URL dial tcp: dial udp 172.25.64.30:53: network is unreachable
 dial tcp: dial udp 172.25.64.30:53: network is unreachable
Get URL dial tcp: i/o timeout

and here's the report output:

Requests        [total]                         71999
Duration        [total, attack, wait]           1h0m0.201463185s, 59m59.952162509s, 249.300676ms
Latencies       [mean, 50, 95, 99, max]         695.300964ms, 256.873692ms, 1.242270337s, 30.775127016s, 30.775127016s
Bytes In        [total, mean]                   4438273450, 61643.54
Bytes Out       [total, mean]                   0, 0.00
Success         [ratio]                         96.68%
Status Codes    [code:count]                    200:69606  505:717  0:1676
Error Set:

I am running vegeta on a 2.6Ghz i7 macbook pro, OSX 10.9.5 do you reckon those errors are caused by me reaching the OS NIC limits?

Thanks

[tsenart]

Hello there,

DNS servers usually run on port 53. Those errors indicate that your DNS server is unavailable for some requests. As a first step, I would attempt to define your targets with IPs instead.

[agenteo]

Great suggestion. I'll try that.

Those DNS errors as well as socket timeout are currently all part of the code 0 (for a total of 1676)? I'd like to contribute if you reckon it's gonna be beneficial to have them broken up in to different classes of codes.

[tsenart]

Those codes are HTTP response status codes. If we're not getting any response it doesn't make sense to classify them as anything else than 0.

[agenteo]

Yeah so maybe code is the wrong name, exception class or kind to be able to see those kind of dns errors bubble up in the report rather then having to search the error set.

I have 1643 entries in my error set from report. But I see 1676 0s codes. Can I get hold of the remaining detail errors?

And I see 717 505s 🐳 which I'd like to investigate on, is there any data on them in the attack output?

Thanks

[tsenart]

For me to help you further, you'd have to provide reproducible steps for me to reach the same erroneous conditions you're seeing.

[agenteo]

would the attack output file be sufficient? Can extract out of it more details then what I see in the vegeta report?

[tsenart]

I also need the inputs to vegeta attack.

[agenteo]

I've been running the same plan off an EC2 instance and also added the -timeout to 30s, fixed all the code 0 problems I was reporting on this.

The 505s were malformed URLs in my test plan will open a separate ticket about that.

[tsenart]

Ignoring errors usually leads to hard to track results. Next time it'd be better for you to do a short load test to validate your inputs, before going full on.