Is there some way to only proxy to non-busy mongrel’s? I’m running 6
mongrels, and if 1 is busy 1 out of every 6 next requests will stall.
The other 5 mongrels could easily handle them quickly. I know the
standard answer for long running tasks is a separate worker process, and
that makes sense for really big long running tasks, but for 10-20 second
or so pages it seems a lot of extra work to split the logic into
multiple pieces if there is an easy way to not send more requests to a
busy mongrel.
Is there some way to only proxy to non-busy mongrel’s? I’m running 6
mongrels, and if 1 is busy 1 out of every 6 next requests will stall.
The other 5 mongrels could easily handle them quickly. I know the
standard answer for long running tasks is a separate worker process, and
that makes sense for really big long running tasks, but for 10-20 second
or so pages it seems a lot of extra work to split the logic into
multiple pieces if there is an easy way to not send more requests to a
busy mongrel.
Apache’s mod_proxy module (and subsequently mod_proxy_balancer) backend
monitoring is AFAIK limited to only being able to see if a proxy
backend is down. It cannot monitor the load or “busy-ness” of the
backend.
I know HA-Proxy and a bunch of hardware balancers does this, I don’t
think nginx, lighttpd and pound does though?
(someone should make a list somewhere with this kinda stuff ;))
So, either switch proxying software, or fix/workaround the original
issue.
On Jan 25, 2007, at 9:40 AM, Johan Sørensen wrote:
multiple pieces if there is an easy way to not send more requests
think nginx, lighttpd and pound does though?
(someone should make a list somewhere with this kinda stuff ;))
So, either switch proxying software, or fix/workaround the original
issue.
JS
Yeah Haproxy or any hardware load balancer can handle this. Apache,
lighty and nginx don’t currently. But Igor, the author of nginx told
me he will add a proxy-max-load directive so you coudl set it to 1
and that mongrel would not get another request until it returns from
the first and the requests would queue in nginx instead of in
mongrel. Probably a week or two before nginx has this built in. This
will give ngixn a significant advantage over apache and lighty.
process, and
backend is down. It cannot monitor the load or “busy-ness” of the
Yeah Haproxy or any hardware load balancer can handle this. Apache,
lighty and nginx don’t currently. But Igor, the author of nginx told
me he will add a proxy-max-load directive so you coudl set it to 1
and that mongrel would not get another request until it returns from
the first and the requests would queue in nginx instead of in
mongrel. Probably a week or two before nginx has this built in. This
will give ngixn a significant advantage over apache and lighty.
Yeah Haproxy or any hardware load balancer can handle this. Apache,
lighty and nginx don’t currently. But Igor, the author of nginx told
me he will add a proxy-max-load directive so you coudl set it to 1
and that mongrel would not get another request until it returns from
the first and the requests would queue in nginx instead of in
mongrel. Probably a week or two before nginx has this built in. This
will give ngixn a significant advantage over apache and lighty.
That’s very cool, Ezra. Do you have any inside knowledge about how
nginx will determine the load of a mongrel? Also, will it work to route
around dead mongrels?
That’s very cool, Ezra. Do you have any inside knowledge about how
nginx will determine the load of a mongrel? Also, will it work to
route
around dead mongrels?
//jarkko
Here is a quote from Igor about this feature:
"I plan to add someting like this:
upstream name {
server host:9000 max_conn=5 max_wait=100 wait_time=60s;
}
"
So for mongrel you could set the max_conn to 1. That would make sure
no mongrel gets another request until it returns from whatever it’s
doing.
Is there some way to only proxy to non-busy
mongrel's? I'm running 6
mongrels, and if 1 is busy 1 out of every 6 next requests will stall.
The other 5 mongrels could easily handle them quickly. I know the
standard answer for long running tasks is a separate worker
process, and
that makes sense for really big long running tasks, but for 10-20
second
or so pages it seems a lot of extra work to split the logic into
multiple pieces if there is an easy way to not send more requests
to a
busy mongrel.
Apache's mod_proxy module (and subsequently
mod_proxy_balancer)
backend
monitoring is AFAIK limited to only being able to see if a proxy
backend is down. It cannot monitor the load or "busy-ness" of the
backend.
I know HA-Proxy and a bunch of hardware balancers does this, I don’t
think nginx, lighttpd and pound does though?
(someone should make a list somewhere with this kinda stuff ;))
So, either switch proxying software, or fix/workaround the original
issue.
JS
Yeah Haproxy or any hardware load balancer can handle this. Apache,
lighty and nginx don’t currently. But Igor, the author of nginx told
me he will add a proxy-max-load directive so you coudl set it to 1
and that mongrel would not get another request until it returns from
the first and the requests would queue in nginx instead of in
mongrel. Probably a week or two before nginx has this built in. This
will give ngixn a significant advantage over apache and lighty.
Thanks. That sounds ideal. I will definitely keep an eye open for this
feature.
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google
Groups "Deploying Rails" group. To post to this group,
send email to [email protected] To
unsubscribe from this group, send email to
[email protected] For more
options, visit this group at
http://groups.google.com/group/rubyonrails-deployment?hl=en
-~----------~----~----~----~------~----~------~--~---
This forum is not affiliated to the Ruby language, Ruby on Rails framework, nor any Ruby applications discussed here.