TCP Connection Reset by Peer

[Shawn Singh]

I'm using Ansible Tower and my Ansible Tower projects are configured to use Git.

From time to time we're seeing situations where project updates are failing.

From the BitBucket access log, I can see the HTTP get requests.

obscured | https | i@1H79XKFx1416x20939365x1 | - | 2020-04-15 23:36:34,512 | "GET /scm/ap/cleanup.git/info/refs HTTP/1.1" | "" "git/1.8.3.1" | - | - | - | - | - | - |

obscured | https | o@1H79XKFx1416x20939365x1 | - | 2020-04-15 23:36:34,516 | "GET /scm/ap/cleanup.git/info/refs HTTP/1.1" | "" "git/1.8.3.1" | 401 | 0 | 0 | - | 3 | 9qddf |

obscured | https | i@1H79XKFx1416x20939367x1 | - | 2020-04-15 23:36:34,517 | "GET /scm/ap/cleanup.git/info/refs HTTP/1.1" | "" "git/1.8.3.1" | - | - | - | - | - | - |

obscured | https | o@1H79XKFx1416x20939367x1 | obscured | 2020-04-15 23:36:34,582 | "GET /scm/ap/cleanup.git/info/refs HTTP/1.1" | "" "git/1.8.3.1" | 200 | 0 | 759 | protocol:1, refs | 65 | 10z49n5 |

obscured | https | i@1H79XKFx1416x20939376x2 | - | 2020-04-15 23:36:34,729 | "GET /scm/ap/cleanup.git/info/refs HTTP/1.1" | "" "git/1.8.3.1" | - | - | - | - | - | - |

obscured | https | o@1H79XKFx1416x20939376x2 | - | 2020-04-15 23:36:34,730 | "GET /scm/ap/cleanup.git/info/refs HTTP/1.1" | "" "git/1.8.3.1" | 401 | 0 | 0 | - | 1 | 2i3fdf |

obscured | https | i@1H79XKFx1416x20939377x2 | - | 2020-04-15 23:36:34,732 | "GET /scm/ap/cleanup.git/info/refs HTTP/1.1" | "" "git/1.8.3.1" | - | - | - | - | - | - |

obscured | https | o@1H79XKFx1416x20939377x2 | obscured | 2020-04-15 23:36:34,755 | "GET /scm/ap/cleanup.git/info/refs HTTP/1.1" | "" "git/1.8.3.1" | 200 | 0 | 759 | protocol:1, refs | 22 | 19emmac |

I don't see any errors on the BitBucket server.

The only coincidence I see in this case is the logs were rotated during the window where the SCM update was happening in Tower.

My question is, would log rotation cause BitBucket to pause servicing requests or do anything to cause RST to be sent to Tower?

Thanks!

Radesh