Re: [courier-users] Re: Automatically use secondary MX host if primary

Jay Lee wrote:

> Rodrigo Severo wrote:
>
>> Ben Kennedy wrote:
>>
>>> Maybe on a 4xx it would make sense. I can't see how it would with a
>>> 5xx,
>>> though. If you have 2 MXs and they don't agree on who is a valid user
>>> (for example), you have bigger problems.
>>>
>>>
>> My fault. You are right. There is no good reason to try a second MX
>> if I get a 5xx error.
>>
>> There still remains the 4xx error situation and the "Connection timed
>> out" => "deferred" situation. Why not try a second MX in these
>> situations?
>>
>>> If it's "turned off", then Courier would try and fail and move on to
>>> the
>>> next one, as it currently does, so what's the difference?
>>>
>> Yes you are right again. I mentioned this because I had this problem
>> with a very old Courier version I was running some time ago. Old
>> problem already solved.
>>
>> But again ther still remains the server too busy situation where I
>> end getting a "Connection time out". Wouldn't it be good to try a
>> second MX here?
>
>
> It's my understanding that Courier would defer the message for awhile
> after the 4xx (which is exactly what a 4xx error is asking it to do)
> and then re-evaluate the MX records. So if there were two mx records
> of the same weight, it would eventually try the 2nd one.

And if the two mx records aren't of the same weight, why not try the
other one? As I said before, this alternative behaviour could be optional.

> In other words, I don't believe Courier "caches" the DNS lookup
> between attempts but I could easily be wrong on this. Sam?

If your are right, I believe I already have half my problem solved.
Let's wait for Sam's answer.

>
> Also, I believe a heavy load server should be configured to refuse
> connections when load is high rather than accept only to defer with
> 4xx errors due to overtaxed resources.

I entirely agree with you.

> Courier does this per the MAXDAEMONS and MAXPERIP settings. If your
> pushing your system resources and running into timeouts, this number
> should be lowered.

Please be sure I would do it if my server were getting timeouts because
of to many connections opened to it. The problem here is that a server I
don't manage is presenting this problem. In a better world I would talk
to the sys admin at the other side and everything would be solved. I
tried. I failed. I have to do something here, sadly.

> If the connection is refused rather than deferred, then a well behaved
> server will then try the next MX record.

I think this is the behaviour of Courier. But I have a doubt here: does
Courier tries the NEXT MX record, as in it keeps a list , tries the
first then the second and so on or Courier tries ANOTHER MX record, as
in it asks for the MX records again and randomly picks a MX record?
Could you please clarify Sam?


Rodrigo Severo




-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
_______________________________________________
courier-users mailing list
courier-users@lists.sourceforge.net
Unsubscribe: https://lists.sourceforge.net/lists/.../courier-users