[ 
https://issues.apache.org/jira/browse/AURORA-1808?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Zameer Manji updated AURORA-1808:
---------------------------------
    Description: 
Thermos loses track of double forking processes, meaning on task teardown  the 
daemonized process will not receive a signal to shut down cleanly.

This can be a serious issue if one is running two processes: 
1. nginx which demonizes and accepts HTTP requests.
2. A backend processes that receives traffic from nginx over a local socket. 

On task shutdown thermos will send SIGTERM to 2 and not 1, causing nginx to 
still accept traffic even though the backend is dead. If thermos could also 
send SIGTERM to 1, the task would tear down cleanly.

  was:
Thermos loses track of double forking processes, meaning on task teardown  the 
daemonized process will not receive a signal to shut down cleanly.

This can be a serious issue if one is running two processes: 
1. nginx which demonizes and accepts HTTP requests.
2. A back and processes that receives traffic from nginx over a local socket. 

On task shutdown thermos will send SIGTERM to 2 and not 1, causing nginx to 
still accept traffic even though the backend is dead. If thermos could also 
send SIGTERM to 1, the task would tear down cleanly.


> Thermos executor should send SIGTERM to daemonized processes 
> -------------------------------------------------------------
>
>                 Key: AURORA-1808
>                 URL: https://issues.apache.org/jira/browse/AURORA-1808
>             Project: Aurora
>          Issue Type: Bug
>            Reporter: Zameer Manji
>            Assignee: Zameer Manji
>
> Thermos loses track of double forking processes, meaning on task teardown  
> the daemonized process will not receive a signal to shut down cleanly.
> This can be a serious issue if one is running two processes: 
> 1. nginx which demonizes and accepts HTTP requests.
> 2. A backend processes that receives traffic from nginx over a local socket. 
> On task shutdown thermos will send SIGTERM to 2 and not 1, causing nginx to 
> still accept traffic even though the backend is dead. If thermos could also 
> send SIGTERM to 1, the task would tear down cleanly.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to