Mailing-List: contact users-help@trafficserver.apache.org; run by ezmlm
Precedence: bulk
Reply-To: users@trafficserver.apache.org
Received-SPF: error (nike.apache.org: local policy)
MIME-version: 1.0
Content-transfer-encoding: 7BIT
Content-type: text/plain; CHARSET=US-ASCII
Subject: Re: host database broken
From: James Peach <jpeach@apache.org>
In-reply-to: <520554AC.9080604@thelounge.net>
Date: Fri, 09 Aug 2013 14:32:00 -0700
Message-id: <35D99FE5-ACB7-4F18-884D-331B2F295490@apache.org>
References: <5204AF1F.40903@thelounge.net>
 <1653730009.20130809123744@thought-mesh.net> <520554AC.9080604@thelounge.net>
To: users@trafficserver.apache.org

On Aug 9, 2013, at 1:44 PM, Reindl Harald <h.reindl@thelounge.net> wrote:

> 
> 
> Am 09.08.2013 19:37, schrieb Alan M. Carroll:
>> Friday, August 9, 2013, 3:58:07 AM, you wrote:
>> 
>>> i am pretty sure that "3.3.5-dev" has the same problem as currently 3.2.5
>>> and that is the cause for "Unable to locate the server named"
>> 
>> I was able to replicate this issue by changing the user ID of the traffic_server process. There is a hidden file that stores some additional HostDB information and if that can't be written the host.db never gets successfully reconfigured. Once I removed the hidden file and made sure the directory permissions matched the ATS user, the host.db was reconfigured successfully. Removing host.db caused a single start to reconfigure and subsequent restarts had no issue.
>> 
>> Check the file etc/trafficserver/internal/hostdb.config and the containing directory to see if that is the problem. In particular that both are owned by 'ats' and not 'nobody'. However, when I did this I got a
> message in diags.log about the permission denied on that file.
> 
> /etc/ is *readonly*
> *period*
> 
> yes, i know that TS complains about read-only filesystem
> that is a bug and should be fixed
> 
> *any* service which *writes* in /etc is broken
> *period*
> 
> services has to store their dynamic data in /var/lib/servicename
> and/or /varcache/servicename and not elsewhere

Yes, I agree that this behaviour can be undesirable. Maybe it's better for your environment to configure trafficserver with --sysconfdir=/var/lib/trafficserver.

If you file a bug, we can look into altering the default behaviour but it's likely to be a difficult task.

> 
>>> -rw-r--r-- 1 ats ats 12M 2013-08-09 10:51 /var/cache/trafficserver/host.db
>> 
>>> * if i delete this file and start TS it get's generated
>>> * but it never works
>>> * the errors below are at every restart
>>> * 3.3.5-dev just don't log it
>> 
>>> if i pull "/var/cache/trafficserver/host.db" from a working host
>>> with 3.2.x with the same config it works
>> 
>>> why can this thing not be disabled at all and normal DNS requests/memory
>>> caching be done like in any other application?

I believe that the reason ATS does not use the system DNS API is that it is a synchronous API and does not support additional records types (eg. SRV records). We would like to clean up (and possibly remove) the HostDB implementation, but currently there is no-one volunteering to do that work, https://cwiki.apache.org/confluence/display/TS/HostDBandDNS.

J