[ 
https://issues.apache.org/jira/browse/TS-1492?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13465755#comment-13465755
 ] 

Leif Hedstrom commented on TS-1492:
-----------------------------------

So, I'm thinking about the feature to not start accepting connections until 
e.g. the cache is initialized, and it would have this problem as well. I'm 
thinking what we should do is to have a stats Record, which is either 0 or 1. 1 
Means that the system is expected to accept() connections properly, and 0 means 
that it is not (e.g. not initialized, or we are throttling).

traffic_cop can read this Records stat before issuing a health check. If it's 
0, then we will not test and restart traffic_server. Maybe with some 
(configurable) threshold, such as that if this setting is at 0 for longer than 
120seconds, then forcefully restart traffic_server.

I think this would solve both of these problems, and becomes a generally useful 
stats that something can check to make sure traffic_server is in a state of 
accepting connections.

                
> if being net_connections_throttled,  ts must to be restarted?(because of 
> can't accept health checking request)
> --------------------------------------------------------------------------------------------------------------
>
>                 Key: TS-1492
>                 URL: https://issues.apache.org/jira/browse/TS-1492
>             Project: Traffic Server
>          Issue Type: Improvement
>          Components: Core
>    Affects Versions: 3.2.0
>            Reporter: kuotai
>            Assignee: kuotai
>            Priority: Critical
>             Fix For: 3.3.0
>
>         Attachments: backdoor_not_throttling.patch
>
>
> In our env, ts will be throttled because of many request incomming 
> simultaneously(because of frontend haproxy is restarted). But we don't expect 
> ts be restarted because of this case. we can handled this:
> 1、if throttled, ts's health check request always be handled
> 2、many many connection request may be handled in long time

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to