Hello,
I have a rails project going with SCGI running to speed it up.
This normally works fine and is very quick compared to rails on its own.
But there are times when I may be try (during testing) to upload or
download very large files and it just stops working dead. Apache seems
to times out (after about 5 min), but for some reason the SCGI process
never seems to recover, or time out, and it brings down the whole
system.
Is there a way of setting a timeout in Zed S.s SCGI config? so that it
just times out that connection instead of bring it all down.
On another point, not far from this one, has anyone set up a cluster of
scgi process and integrated them into apache with some sort of load
balancing method at all? Any examples?
Cheers.
Dan.
Fish Man wrote:
Hello,
I have a rails project going with SCGI running to speed it up.
This normally works fine and is very quick compared to rails on its own.
But there are times when I may be try (during testing) to upload or
download very large files and it just stops working dead. Apache seems
to times out (after about 5 min), but for some reason the SCGI process
never seems to recover, or time out, and it brings down the whole
system.
Is there a way of setting a timeout in Zed S.s SCGI config? so that it
just times out that connection instead of bring it all down.
I’m not sure if there is a way to set it in Apache. In lighttpd, its
“disable-time” => 0.
On another point, not far from this one, has anyone set up a cluster of
scgi process and integrated them into apache with some sort of load
balancing method at all? Any examples?
Install the most current scgi and ruby-style gems. You can set a
cluster to run on multiple ports or a single port. The web server
configuration is the exactly same with single port clustering as it is
with running a single scgi listener. With either type of clustering,
you can restart without losing requests. Read the README that comes
with the ruby-style gem for details.
I took over the SCGI project from Zed S., so feel free to ask me any
questions about it.
Jeremy
Jeremy E. wrote:
Install the most current scgi and ruby-style gems. You can set a
cluster to run on multiple ports or a single port. The web server
configuration is the exactly same with single port clustering as it is
with running a single scgi listener. With either type of clustering,
you can restart without losing requests. Read the README that comes
with the ruby-style gem for details.
I took over the SCGI project from Zed S., so feel free to ask me any
questions about it.
Jeremy
Hi thanks for your advice.
I think that I understand how to start up clusters for scgi with
scgi_cluster, that seems to be well documented.
But im not exactly sure what do I need to put into Apache to distribute
the load to each of the different processes. There seems to be a good
example for lighthttd is there an example avalable for apache2?
Cheers.
Dan.
Fish Man wrote:
Jeremy E. wrote:
Install the most current scgi and ruby-style gems. You can set a
cluster to run on multiple ports or a single port. The web server
configuration is the exactly same with single port clustering as it is
with running a single scgi listener. With either type of clustering,
you can restart without losing requests. Read the README that comes
with the ruby-style gem for details.
I took over the SCGI project from Zed S., so feel free to ask me any
questions about it.
Jeremy
Hi thanks for your advice.
I think that I understand how to start up clusters for scgi with
scgi_cluster, that seems to be well documented.
But im not exactly sure what do I need to put into Apache to distribute
the load to each of the different processes. There seems to be a good
example for lighthttd is there an example avalable for apache2?
scgi_cluster hasn’t worked since 0.4.3, and that version doesn’t support
single-port clustering, which seems to be a requirement if you want to
use SCGI with Apache. As I said earlier, install the most current scgi
and ruby-style gems, and read the README that comes with ruby-style. It
has deployment instructions as well as a sample Apache configuration.
Jeremy
Its seems that Mongrel is easier to set up and configure. Out of
interest how does Mongrel compare to SCGI, in terms of speed?
Sorry That was in response to Fish Man,
“Its seems that Mongrel is easier to set up and configure. Out of
interest how does Mongrel compare to SCGI, in terms of speed?”
I have not been able to test the new Mongrel much but as of yet I feel
that a mongrel cluster running behind any reverse proxy runs faster
than using SCGI.
It is also much easier to setup.
-Ron
Fish Man wrote:
Its seems that Mongrel is easier to set up and configure. Out of
interest how does Mongrel compare to SCGI, in terms of speed?
A cluster of SCGI listeners is the same speed as a cluster of Mongrel
listeners, the difference is not statistically significant. I don’t
think it is any more difficult to set up, but there is more
documentation about how to set up a mongrel cluster. If you read the
README that comes with ruby-style, you’ll see setting up a cluster of
SCGI listeners is quite simple, but that’s the only place you’ll find
documentation.
Jeremy
One thing to note is RAM usage. Each mongrel instance requires about
30-40MB ram because it spwns its own ruby process. If you are running
on a VPS or something similar you may need to rethink your deployment
strategy.
-Ron