Re: Count retries and redispatches

On Fri, 19 Oct 2007, Willy Tarreau wrote:

> Hi Krzysztof,
>
> On Fri, Oct 19, 2007 at 12:01:16AM +0200, Krzysztof Oledzki wrote:
>> I was trying to find what exactly did you change, unfortunately gitweb
>> from 1wt does not support commitdiff (instead of diff I get a list of old
>> and new files).
>
> It's an old (and painfully slow) version. I hate it, sometimes it works,
> sometimes it breaks. It runs on an old slow HPPA machine at 60 MHz over
> NFS, and I guess it's one of the reasons why it behaves so poorly.

;)

>> I was also trying to download full git repository with:
>>
>> git clone http://git.1wt.eu/haproxy-1.3.git haproxy-1.3.git
>>
>> but I ended with:
>>
>> error: The requested URL returned error: 502 (curl_result = 22, http_code =
>> 502, sha1 = 3706cecc373ca19dbef788bbcc08306d59300598)
>
> There's no reason for this to happen unfortunately :-(
> And for once, I do not have traces of oopses in the dmesg on the
> server. I've checked my local logs and I don't see any 502, so may
> be it happened somewhere between us because the server was too slow
> to respond ?

Quite likely as downloading 10MB takes _very_ long time. I don't know if it is possible to change timeouts in git clone, any chances for supporting the "git://" protocol?

>> Getting pack list for http://git.1wt.eu/haproxy-1.3.git
>> Getting index for pack 52ae90d594ec7b801faf5508d8e67c7467e6b457
>> Getting index for pack 79486d69b47bab3775a19e73d70d7767646c648f
>> Getting pack list for http://git.1wt.eu/haproxy-1.2.git/
>> error: Unable to find 3706cecc373ca19dbef788bbcc08306d59300598 under
>> http://git.1wt.eu/haproxy-1.3.git
>> Cannot obtain needed tree 3706cecc373ca19dbef788bbcc08306d59300598
>> while processing commit 8d9246d282218422ccfa005731c6e0b2ff6cf75f.
>>
>> Anyway, comparing manually downloaded files from
>> http://haproxy.1wt.eu/git/?p=haproxy-1.3.git;a=commitdiff;h=1cf36ba3ae9e532ac06aaf2a30e6c217d8d1f878
>> shows that:
>>
>> --- 3 Thu Oct 18 23:43:05 2007
>> +++ 4 Thu Oct 18 23:43:27 2007
>> @@ -574,6 +574,10 @@
>> {
>> /* we are in front of a retryable error */
>> t->conn_retries--;
>> + if (t->srv)
>> + t->srv->retries++;
>> + t->be->retries++; <<<--------- this one
>> +
>> if (t->conn_retries < 0) {
>> /* if not retryable anymore, let's abort */
>> tv_eternity(&t->req->cex);
>
> (...)
>
>> This is exactly what I send. So, it seems that if there is a fix needed it
>> may be missing.
>
> Dman, you're right! The line I had to add by hand was the line
> I marked above [ t->be->retries++ ]. But re-reading your mail, I see
> it was already there. I had to apply your patch by hand due to the
> numerous changes I did in the stats area. So most likely I have myself
> jumped over it when copy-pasting parts.

Right, I must be very ugly to apply patch by hand. Next time I would try to base on the git repository if only I would be able to clone it. :(

> Thanks for noticing, and please accept my apologizes for mistakenly
> making it look like your patch wasn't perfect !

Really, not need to. I was just curious what I overlooked this time (like in the spread-checks patch) and was quite surprised that nothing was changed. ;)

Best regards,

                                 Krzysztof Olêdzki Received on 2007/10/19 00:32