the version 1.0.0 of dfget still have zombie processes after pull image

[liy516]

phenomenon

[root@dragonfly01 dragonfly]# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
99b11e0b94f0 192.168.56.102/common/dfclient:1.0.0 "/opt/dragonfly/df-c…" 22 minutes ago Up 6 minutes 0.0.0.0:65001->65001/tcp dfclient
[root@dragonfly01 dragonfly]# ps -ef | grep dfget
root 5039 4919 0 10:58 ? 00:00:00 [dfget]
root 5052 4919 0 10:58 ? 00:00:00 [dfget]
root 5057 4919 0 10:58 ? 00:00:00 [dfget]
root 5063 4919 0 10:58 ? 00:00:00 [dfget]
root 5604 2456 0 11:05 pts/0 00:00:00 grep --color=auto dfget

[fengzixu]

Can you fill the re-produce method in this issue? I think it is beneficial for investigating the problem.

[zcc35357949]

https://github.com/dragonflyoss/Dragonfly/blob/master/dfget/core/uploader/peer_server_executor.go#L75
https://github.com/dragonflyoss/Dragonfly/blob/master/dfget/core/uploader/peer_server_executor.go#L79
Locking after checking server seems to lead this.

[liy516]

Can you fill the re-produce method in this issue? I think it is beneficial for investigating the problem.

Pulling a new mirror produces the same number of zombie processes as the number of levels of the mirror.

[liy516]

https://github.com/dragonflyoss/Dragonfly/blob/master/dfget/core/uploader/peer_server_executor.go#L75
https://github.com/dragonflyoss/Dragonfly/blob/master/dfget/core/uploader/peer_server_executor.go#L79
Locking after checking server seems to lead this.
You have submitted a PR, I try to compile your code and test it.