(ConnectionFailure Network.BSD.getProtocolByName: does not exist (no such protocol name: tcp))) when trashing connections

[reactormonk]

Point at a bucket with a lot of small files. I got a (ConnectionFailure Network.BSD.getProtocolByName: does not exist (no such protocol name: tcp))) pretty reliably.
https://gist.github.com/reactormonk/8106134ccb59190b9cdf8eedd2b66cff

Fixing the issue by moving the sampling one back: https://gist.github.com/reactormonk/8106134ccb59190b9cdf8eedd2b66cff/revisions

batch-s3: TransportError (HttpExceptionRequest Request {
  host                 = "s3.eu-central-1.amazonaws.com"
  port                 = 443
  secure               = True
  requestHeaders       = <censored>
  path                 = "<censored>"
  queryString          = ""
  method               = "GET"
  proxy                = Nothing
  rawBody              = False
  redirectCount        = 0
  responseTimeout      = ResponseTimeoutMicro 70000000
  requestVersion       = HTTP/1.1
}
 (ConnectionFailure Network.BSD.getProtocolByName: does not exist (no such protocol name: tcp)))

[snoyberg]

Can you provide a repro that uses only the http-client and http-client-tls libraries?

[rsarres]

I had this problem running a Haskell app on an Ubuntu docker image. Installing build-essential package solved the problem.

[kingxueyuf]

Having the same issue as well

[snoyberg]

No minimal repro, and this is likely a file descriptor issue from the sound. Closing.

[reactormonk]

Yeah, the code runs out of file descriptors. See #307

[exarkun]

I had this problem running a Haskell app on an Ubuntu docker image. Installing build-essential package solved the problem.

FWIW, you only need netbase to fix the problem in that context (12.7 KiB compared to 21.7 MiB for build-essential).

[creichert]

I believe @exarkun is right. I had this issue recently and had to install netbase. What's odd is that the failures seemed intermittent.

If the issue persists after installing netbase, check the /etc/hosts file of the to make sure it's sane. I believe at minimum:

$ cat /etc/hosts
127.0.0.1       localhost
::1             localhost ip6-localhost ip6-loopback
ff02::1         ip6-allnodes
ff02::2         ip6-allrouters

[ketzacoatl]

Installing netbase fixed this for me, build-essential wasn't enough. Without netbase, I was unable to use cache-s3 or stack build (and ghc installs failed). It was a mess. After installing netbase, caching and builds worked fine. Does anyone know why or how netbase fixes this problem?

[bradrn]

I was getting this same error on a Yesod app when deployed to AWS Elastic Beanstalk. Installing netbase fixed it for me as well.

[tomjaguarpaw]

Specifically, the file that seems to be required is /etc/protocols

[tomjaguarpaw]

Also seems to require /lib/x86_64-linux-gnu/libnss_files.so.2.