1

I am running a process in a docker container that periodically does a git clone via https from gitlab. Most of the time it works, but every so often it stalls indefinitely and requires me to manually destroy and recreate the docker container to fix. When I enable settings GIT_TRACE=1 GIT_CURL_VERBOSE=1 and --verbose, I can see in the logs:

trace: run_command: 'git-remote-https' 'origin' 'https://myusername:mypassword@gitlab.com/username/reponame.git'
* Couldn't find host gitlab.com in the .netrc file; using defaults
* About to connect() to gitlab.com port 443 (#0)
*   Trying xx.xxx.xxx.xx...
* Connected to gitlab.com (xx.xxx.xxx.xx) port 443 (#0)
* Initializing NSS with certpath: sql:/etc/pki/nssdb
*   CAfile: /etc/pki/tls/certs/ca-bundle.crt
  CApath: none

I'm most confused about the inconsistent nature of this error. Per this other thread I tried to set git config http.postBuffer 524288000 with no luck. How should I continue to troubleshoot? Thank you

chris.mclennon
  • 966
  • 10
  • 25

1 Answers1

1

It turns out this issue was on a different layer entirely, unrelated to any of the tagged technologies. This was an error based on MTU configurations. The machine I was running the git process on had an MTU value that was very close to a router's accepted MTU value. After adding the size of the packet plus its packet overlay, the final packet size exceeded the router's accepted MTU value and therefore dropped the packet. The resolution was to raise the router's accepted MTU value.

chris.mclennon
  • 966
  • 10
  • 25