On 10/5/2010 8:56 PM, William A. Rowe Jr. wrote:
On 10/5/2010 5:41 PM, Daniel Ruggeri wrote:
All;
With the talk about a 2.2.17 coming soon, I would very much like to get the
remaining
requisite votes and implementation of the patch (48939 - in STATUS currently) I
had
submitted for inclusion. I know a lot of folks are rather busy these days, but
I was
hoping I could draw attention to this again in hopes of making the 2.2.17
release.
P.S.
I would love to include details of this patch in my ApacheConNA 2010
session as it
helps address some of the shortfalls the intelligence shortfalls.
Just as a suggestion, most of us don't memorize numbers (... for example, I
can't
remember my own kids cell phone numbers, my phone does so for me.)
So when someone want eyeballs on an issue, please remind us the subject, and if
it is not too lengthy, attach the patch. Consider that sometimes our chance to
react to your email is in the air, devoid of network access, and we are just
trying to plow through our email queue offline.
All that said, trawick, niq and wrowe have all reviewed this specific backport,
and it is in the queue to be applied to 2.2.
William;
Understood - I was too busy repeating myself in the last sentence I
didn't think to provide more details. I also must have misread STATUS
when I checked on this the other day. Thank you for the response.
On a different note, I recall you brought the topic up about worker
acquiescence in a planned maintenance situation. I am not sure if folks
had a chance to review what I brought up, but I have submitted a patch
to do this. However, I would really prefer input on the patch as I am
not 100% sure it is ready for proposal in STATUS. Also because,
technically, one could set the redirect route for the worker and force
its traffic elsewhere (works fine in a two node situation, but distorts
load distribution if there are more).
Bug URL
https://issues.apache.org/bugzilla/show_bug.cgi?id=48841
Patch notes:
I used a constant called PROXY_WORKER_NOLBFACTOR in mod_proxy.h and
changed
the atoi call during configuration to strtol since atoi. I did this
because the
atoi call returns 0 both during error situations and when the proper
value to
return is 0. Also, the existing checks had to be refactored a little
since (at
least on the SUN c compiler) an uninitialized integer is the same as
`0'. Aside
from that, only the bybusiness algorithm had to be modified to avoid a
divide
by zero error.
--
--
Daniel Ruggeri
