Re: [PATCH] : Prevent redispatcher from selecting the same server

From: Krzysztof Oledzki <ole#ans.pl>
Date: Fri, 28 Dec 2007 13:39:17 +0100 (CET)

On Wed, 26 Dec 2007, Willy Tarreau wrote:

> On Wed, Dec 26, 2007 at 03:29:37PM +0100, Krzysztof Oledzki wrote:
>>>> This tcr is for "transition check rate" so you have a global 1-100%
>>>> paremeter for it, exactly like spread_checks.
>>>
>>> Basically what we need indeed, except that I'd really like to have two
>>> different values and not just a factor between them. While most people
>>> would prefer to use faster checks during transitions, others will prefer
>>> to slow them down in order to give servers a chance to recover.
>>
>> Oh, good point. I didn't think about such scenario. OK, how about making
>> it a local parameter, that cen be set in defaults section and changing
>> range from 1-100% to 1-1000%? IMHO in most cases it is easier to set
>> "transcheckrate 25%" for all backends that adding transcheck for each
>> server.
>
> I'm not fond of applying percentages to timers. It quickly becomes uneasy
> to manipulate. For instance, people will have a hard time trying to figure
> out after how many seconds a server will be seen down when there is a 15s
> inter and 25% transcheckrate with 3 retries :-/

Right.

> I understand your concern about the default value, but it should be addressed
> differently IMHO, as other server parameters are getting annoying too. In fact,
> we simply need a "default server" statement with some reusable values such as
> "inter", "fall", "rise", "fastinter", "check", "port", "source" and probably
> other things, and that one should be per-backend, and should be definable in
> a "defaults" section.
>
> What do you think about this ?

Yes, that makes sense. It should be quite easy to do this, all what we need here is to make sure it is possible to disable for a specific server already per-backend enabled options. I'll cook a RFC patch for both transcheck and a "default server" statement.

Best regards,

                                 Krzysztof Olędzki Received on 2007/12/28 13:39

This archive was generated by hypermail 2.2.0 : 2007/12/28 13:45 CET