madasebrof
Super weird http issue
Wow, this is an odd one.
I’m making an http request (post) to a server.
- Work fine from curl
- Works fine from iex
- Bombs out when I run from within app!
I have literally never had this happen before, when the EXACT same function behaves differently live vs. iex.
Just wondering if anyone has insight into some magic that might be going on behind the scenes that could be causing this! (This is a phoenix app, but doesn’t seem phoenix-related…)
I even added a checksum function to make sure the data being sent was identical & there wasn’t some odd file-write delay causing the issue.
Example - Success from iex (note–iex is actually connected to the running app!):
iex(91be5e1d30ff@172.21.0.3)5> Polo.Ib.Connect.post(body, "/create")
13:24:11.570 [info] Ib.Connect.post endpoint: [ https://qa.xyz.com/ws/eca/create ]
13:24:11.573 [warn]
Mojito.request(
method: :post,
url: https://qa.xyz.com/ws/eca/create,
headers: [{"Content-Type", "application/json"}, {"User-Agent", "Elixir"}, {"Accept", "*/*"}],
body: data -> checksum = 904033a8a73bdea778edf96317957f40704686cd,
opts: [timeout: 180_000, transport_opts: [verify: :verify_none]]
)
{:ok,
%{
"batchId" => 305738,
"fileData" => %{
"data" => "hQ ... eFBJXdc/cKHQ==",
"name" => "sgimFA_20200211-082414031_965716.xml"
},
"isProcessed" => true,
"timestamp" => "2020-02-11 08:24:15"
}}
works great!
From with the app, when running programmatically, I get
[info] Ib.Connect.post endpoint: [ https://qa.xyz.com/ws/eca/create ]
[warn]
Mojito.request(
method: :post,
url: https://qa.xyz.com/ws/eca/create,
headers: [{"Content-Type", "application/json"}, {"User-Agent", "Elixir"}, {"Accept", "*/*"}],
body: data -> checksum = 904033a8a73bdea778edf96317957f40704686cd,
opts: [timeout: 180_000, transport_opts: [verify: :verify_none]]
)
[error] %Mojito.Response{body: "<!DOCTYPE html>\n<html>
\n<head>\n<title>Account Management Unavailable | ... \" content=\"500 Error\">\n<<> ...,
complete: true,
headers: [{"date", "Tue, 11 Feb 2020 13:21:47 GMT"},
{"server", "Apache"},
{"vary", "\\\"Origin\\\""},
{"content-security-policy", "frame-ancestors 'self' ... .com;"},
{"content-length", "5879"},
{"connection", "close"},
{"content-type", "text/html; charset=UTF-8"}],
status_code: 500}
Note that the checksum of the data is identical! The headers are identical! It’s literally the exact same function with identical arguments.
I am just totally at a loss as to what could be different.
Any thoughts are appreciated!
ps Was using HTTPPoision, then switched to Mojito/Mint to see if that would fix, but same issue.
Most Liked
voltone
Decrypting TLS would be possible in two scenarios:
- You have access to the server’s private key and you constrain the negotiated cipher suites to ensure the key exchange does not use (EC)DHE, or
- You write the master secret for the session you want to decode to a file
This last option can work both on the server side (if you control the server) or on the client side (using Erlang’s :ssl.connection_information(socket, [:client_random, :master_secret]).
But there is an easier way: use Erlang’s tracing functionality, e.g. using :dbg or some higher level abstraction such as recon:
iex(1)> :dbg.tracer()
{:ok, #PID<0.258.0>}
iex(2)> :dbg.p(:all, :call)
{:ok, [{:matched, :nonode@nohost, 119}]}
iex(3)> :dbg.tp({Mint.Core.Transport.SSL, :send, :_}, :cx)
{:ok, [{:matched, :nonode@nohost, 1}, {:saved, :cx}]}
iex(4)> {:ok, response} = Mojito.request(method: :get, url: "https://github.com")
(<0.269.0>) call 'Elixir.Mint.Core.Transport.SSL':send({sslsocket,{gen_tcp,#Port<0.6>,tls_connection,undefined},
[<0.303.0>,<0.302.0>]},[[<<"GET">>,32,<<"/">>,<<" HTTP/1.1\r\n">>],
[[[<<>>,<<"content-length">>,<<": ">>,<<"0">>,<<"\r\n">>],
<<"host">>,<<": ">>,<<"github.com">>,<<"\r\n">>],
<<"user-agent">>,<<": ">>,<<"mint/1.0.0">>,<<"\r\n">>],
<<"\r\n">>,<<>>]) ({'Elixir.Mint.HTTP1',request,5})
(<0.269.0>) returned from 'Elixir.Mint.Core.Transport.SSL':send/2 -> ok
{:ok,
%Mojito.Response{...}
It takes a second to spot the actual request data in the output and perhaps reformat it, but it is much easier than setting up TLS decoding ![]()
madasebrof
At the end of the day, I think it’s just a super flaky endpoint, and it was more coincidence that some requests were working and other were not!
I just tried 4 request in a row–first 3 bombed, last one went through.
Again, thanks for all ideas!
Rock on,
Popular in Questions
Other popular topics
Categories:
Sub Categories:
Forums
Popular Tags
- #ecto
- #liveview
- #troubleshooting
- #learning-elixir
- #deployment
- #library
- #erlang
- #testing
- #genserver
- #mix
- #absinthe
- #remote-other
- #otp
- #plug
- #how-to-question
- #macros
- #postgres
- #channels
- #elixirconf
- #exunit
- #discussion
- #code-sync
- #javascript
- #podcasts
- #onsite
- #dialyzer
- #docker
- #authentication
- #umbrella
- #full-time-contract
- #podcasts-by-brainlid
- #ecto-query
- #elixir-ls
- #phoenix_html
- #iex
- #blog-post
- #graphql
- #genstage
- #ai
- #websockets
- #supervisor
- #advent-of-code
- #elixirconf-us
- #distillery
- #processes
- #forms
- #api
- #metaprogramming
- #security
- #performance









