On Sat, 29 Sep 2007, Willy Tarreau wrote:
> Hi Krzysztof,
Hi, :)
> sorry for this long delay, I've been working on other complex things and
> did not want to get my mind polluted by completely different things.
No problem at all, I know this problem very well. ;)
> On Sun, Sep 23, 2007 at 03:29:21PM +0200, Krzysztof Oledzki wrote:
>> Hello,
>>
>> It is important to know how does your instalation perform. Haproxy masks
>> connection errors, which is extremely good for a client but it is bad for
>> an administrator (except people believing that "ignorance is a bless").
>>
>> Attached patch adds retries and redispatches counters, so now haproxy:
>>
>> 1. For server:
>> - counts retried connections (masked or not)
>
> I think that's useful because on the stats page, if you see too many retries,
> it means that you have a connection problem to that server. And it's very
> common in datacenters.
Indeed.
>> 2. For backends:
>> - counts retried connections (masked or not) that happened to
>> a slave server
>> - counts redispatched connections
>> - does not count successfully redispatched connections as backend errors.
>> Errors are increased only when client does not get a valid response,
>> in other words: with failed redispatch or when this function is not
>> enabled.
>
> This seems fair enough. The doc will need some lifting to explain that,
> because it's somewhat different to how it's done today.
Exactly.
>> 3. For statistics:
>> - display Retr (retries) and Redis (redispatches) as a "Warning"
>> information.
>
> How about the logs ? We may add a flag indicating if the connection has been
> redispatched or not, and if it got some retries or not. We don't need to log
> the number, just the fact that it happened. This helps understanding the
> timers.
OK, I will think about it and also try to create some patches for the documentation for all my accepted changes.
>> I think it is a post-1.3.12 material.
>
> Yes, I'm queuing it for integration.
Thank you.
Best regards,
Krzysztof Olędzki Received on 2007/09/29 17:43
This archive was generated by hypermail 2.2.0 : 2007/11/04 19:21 CET