[
https://issues.apache.org/jira/browse/TS-4220?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15168940#comment-15168940
]
Alan M. Carroll commented on TS-4220:
-------------------------------------
I'm not seeing the problem here. {{failover_now}} checks if the DNS server has
been unresponsive for too long, returning {{true}} if that is the case, meaning
the nameserver is unavailable for Traffic Server. In that case {{rr_failure}}
should be called because that is the method that updates the internal state to
mark that nameserver as down. Further, changing the sense of the check would
mean doing a retry immediately when the nameserver is marked down, which
doesn't seem to be the intent the just prior {{for}} loop is run to do such
retries after some time,{{DNS_PRIMARY_RETRY_PERIOD}}, has gone by.
It seems to me the current logic is therefore correct - call {{rr_failure}} if
{{failover_now}} returns {{true}}. You can see the same logic in the non-round
robin case a few lines below.
> DNSHandler::mainEvent
> ---------------------
>
> Key: TS-4220
> URL: https://issues.apache.org/jira/browse/TS-4220
> Project: Traffic Server
> Issue Type: Bug
> Reporter: song
>
> Dear all
> failover_now return false when the name server's mark is not expiring, and
> we need to call rr_failure to mark the name server as down.
> Thx!!
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
