So this is really for the Googles. The problem turned out to be my code
(isn’t it always) and a misuse of an edge triggered epoll loop. Sorry for
all the line noise list.
Fred
On 1/15/15, 2:06 PM, "Luke Bakken" wrote:
>Most reverse proxies have a timeout for data being sent from the
>server.
So it does not take much time. Less then a second. Here is a little bit
more data. So it looks like (after increasing the buffer windows on both
linux hosts) I get the first connection to fully complete. And then it
looks like riak stops responding to new requests.
I increased the backlog for we
Most reverse proxies have a timeout for data being sent from the
server. If this query takes a long time to run when not using the
proxy, I suggest investigating increasing timeout values.
--
Luke Bakken
Engineer
lbak...@basho.com
On Thu, Jan 15, 2015 at 12:01 PM, Fred Grim wrote:
> The more I l
The more I look at this the more I am thinking there is something busted
in riak or webmachine (I am using riak 1.4.10). The issue seems to be
impervious to apache settings or haproxy settings and going through the
source there doesn¹t seem to be many knobs to turn to adjust chunked
responses. Goog
So actually yeah and the problem was the same. However if I hit riak
directly then there is no issue. Are there http settings for riak that
cause the streaming behavior to change when being proxied?
F
On 1/14/15, 10:31 PM, "Toby Corkindale" wrote:
>Just a thought, but have you tried to replica
Just a thought, but have you tried to replicate the problem when using
haproxy (with the example configs provided on basho's site)?
That might let you discover if the problem really IS with Apache, or
if it's something to do with your Riak config instead.
On 15 January 2015 at 11:05, Fred Grim wr
Hello List,
So I am using apache on ubuntu as a reverse proxy to a riak cluster. Everything
seems to work fine except for largish streaming map reduce queries. The config
is:
ServerAdmin aper...@acompany.com
BalancerMember http://:8098
BalancerMember http://:8098
