解决Gitlab的The remote end hung up unexpectedly错误

Gitlab服务器环境是CentOS7＋Gitlab7.2.1，最近发现在开发机上使用git pull更新文件时，会报如下错误。

fatal: The remote end hung up unexpectedly
fatal: early EOF
fatal: unpack-objects failed

使用git clone重新checkout源也受到影响，长时间checkout不出来，

Google了半天也没找到个好办法，最后还是查错误日志定位到了问题

在/var/log/gitlab/unicorn/unicorn_stderr.log中，发现如下的错误信息

E, [2014-12-06T09:13:10.319216 #11074] ERROR -- : worker=0 PID:11091 timeout (31s > 30s), killing
E, [2014-12-06T09:13:10.336186 #11074] ERROR -- : reaped #<Process::Status: pid 11091 SIGKILL (signal 9)> worker=0
I, [2014-12-06T09:13:10.340379 #11183]  INFO -- : worker=0 spawned pid=11183
I, [2014-12-06T09:13:10.340848 #11183]  INFO -- : worker=0 read

看来是被Ruby误认为超时中断了。解决办法就是调大unicorn的timeout值。

修改/var/opt/gitlab/gitlab-rails/etc/unicorn.rb，将

# What the timeout for killing busy workers is, in seconds
timeout 30

改为

# What the timeout for killing busy workers is, in seconds
timeout 60

最后，运行

sudo systemctl restart gitlab-runsvdir.service 

重启Gitlab。

之后客户端就可以正常Git pull或者clone了