[
https://issues.apache.org/jira/browse/TS-4436?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15285030#comment-15285030
]
ASF subversion and git services commented on TS-4436:
-----------------------------------------------------
Commit 5f0a649f7af0f6616b81d577a635e6997344b9e1 in trafficserver's branch
refs/heads/master from [~jacksontj]
[ https://git-wip-us.apache.org/repos/asf?p=trafficserver.git;h=5f0a649 ]
TS-4436: Move hosts file implementation to `do_dns`
This moves the hosts file overrides down to the DNS layer instead of the
HostDBInfo layer. This means that each port will get its own entry in hostdb,
which will exist as its own HostDBInfo just like all the rest of the entries
(with a TTL of the time remaining until the next hostdb sync). This also means
we only have to do the check in the map on DNS lookup instead of on each
probe().
In addition to fixing the down status issues, this also means we no longer need
to keep old copies of the strings etc. since they are copied once lookup_done
is called.
> ATS' tracking of host failures is broken for entries acquired through the
> hosts file
> ------------------------------------------------------------------------------------
>
> Key: TS-4436
> URL: https://issues.apache.org/jira/browse/TS-4436
> Project: Traffic Server
> Issue Type: Bug
> Components: HostDB
> Reporter: Thomas Jackson
> Assignee: Thomas Jackson
> Fix For: 7.0.0
>
>
> The current implementation of the hosts file maintains a map of name to
> hostdbinfo, which means that any lookups that come into the probe method
> return the same hostdbinfo struct. This means that any requests to a given
> address (regardless of port) recieve the same HostDBInfo struct-- which in
> turn means that the up/down status is global to the IP instead of per ip/port
> pair.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
