Ajax really slow in production, very weird

Rails
1

It works fine when I start a single instance of mongrel or thin. But
when I use thin cluster or passenger, some ajax is really slow, takes
about 2 minutes to complete. Any sugestion? By the way, I tried
passenger for nginx and nginx + thin cluster.

2

What does your production log say? How many ms does it take to render
your request?

3

It looks that the request has problems reach a thin instance, as soon
as it reaches the thin instance, the request completes quickly. The
production log says it took about 150000ms to complete the request.
But not all the request has the problem, some of them complete within
half a second which is normal. I’m guessing that nginx has problems
dispatching the request to a thin or passenger instance. Has any body
encountered this problem before?

On Oct 26, 1:22 am, Wouter de Bie [email protected]

4

first, I’m sure my connection to the internet is ok.
And the problem exist after i restart nginx and thin, and it is
repeatable.
I did tested on my local machine with the same configuration, it works
fine.
I used ubuntu 8.10 on my local machine, but the server runs on freebsd
7.2, so maybe there’s some difference.

5

On Sun, Oct 25, 2009 at 6:28 PM, phoenix [email protected] wrote:

It looks that the request has problems reach a thin instance, as soon
as it reaches the thin instance, the request completes quickly. The
production log says it took about 150000ms to complete the request.
But not all the request has the problem, some of them complete within
half a second which is normal. I’m guessing that nginx has problems
dispatching the request to a thin or passenger instance. Has any body
encountered this problem before?

Is this a local or remote deployment? In any case, I would do the
following:

  1. check your connection to the internet
  2. check with your New Relic or other performance monitoring to track
    the
    issue
  3. re-check the overall configuration of the system
  4. reboot nginx and thin servers and see if the problem still persists
  5. check with the hosting company
  6. do testing in production mode on you local system using the same
    remote
    configuration to see if the issue exists

Good luck,

-Conrad

6

I got the problem solved!

It turned out to be a directory access problem. I used texlive for
math formular editing, and the access mod for directory that contains
these executables is 777. After I changed it to 755, the problem
disapeared. I overlooked a warning in my log, which says: ‘world
writable directory /xxx/xxx’. And I didn’t expect this would cause
such weird behavior, I don’t see much connection here…

I wasted a lot of time on this and it really taught me a lesson!

Thanks for all your kind help. You are a very nice man!

7

On Sun, Oct 25, 2009 at 7:30 PM, phoenix [email protected] wrote:

first, I’m sure my connection to the internet is ok.

You’re sure but did you check it.

And the problem exist after i restart nginx and thin, and it is
repeatable.

I did tested on my local machine with the same configuration, it works

fine.

I used ubuntu 8.10 on my local machine, but the server runs on freebsd

7.2, so maybe there’s some difference.

It’s not really the same configuration if you’re using binaries built
for
Linux
locally and FreeBSD binaries remotely. In theory, they would be the
same
but in practice they may not be. Did you check with your hosting
company?
What do the logs say for both nginx and thin? What’s the performance
like
without Javascript enabled?

-Conrad