Hi Tres, Thanks for the reply. I have provided a couple of notes below, and I am copying to the list: Quoting Tres Seaver <tseaver@zope.com>:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
David Swift wrote:
| Running Zwiki 0.36 on Zope 2.7.4-0 on Python 2.4, glibc-2.3.4-3, kernel | 2.4.22-nptl (fedora core 1).
Python 2.3.4 is the recommended, supported version of Python.
Understood - we were running on Python 2.3.3, using Zope 2.7.0 on a 2.4.18 kernel previously. I had hoped to take advantage of bugfixes in threading on Python 2.4. Nevertheless, since I also see this delay using Python 2.3.3 on the new machine, I am pretty certain it is not a problem with Python. I am also pretty certain it is not Zope 2.7.4 per se, as I see the same behavior when using 2.7.3 on the new machine. I really suspect I have left out some critical piece of configuration, but I am lost as to what that may be.
| Whenever we hit a ZWiki page, there is a delay of 8 seconds before the | first page is sent. Subsequent objects for that request are sent in | reasonable (i.e. sub-second) times. Every individual page request seems | to cause this delay. I have tried all 3 ways of setting the | LD_ASSUME_KERNEL variable, tried recompiling Python using the | '--with-lib=' option for the different thread libraries, tried Python | 2.3.3, and all this still gives the same result - a nasty 8 seconds delay.
It looks as though your delay is in communicating with a ZEO server -- the speedup after first load would be explained by the presence of the object in the ZEO client cache.
We are not using a ZEO server, we are still on filestorage. I wish I could use ZEO, but when I first set it up, I think I hit this problem as well, and I went back to filestorage. I am also pretty certain that the pages are not being cached. Every time I click on a ZWiki link, or click the 'FrontPage' link at the top of a ZWiki page, we experience the delay. As I said, every individual page request has this delay, no matter how many times the pages have been loaded previously. Looking into the detailed cache debug info for the instance, I never see a ZWiki page with a count higher than 1.
| Has anyone else run into a delay like this? Is anyone using a similar | kernel/Python/Zope combination successfully?
I run Zope 2.7.4 with ZEO (on the same machine) on FC1 without seeing such issues, using a Python built for the purpose (I don't reuse /usr/bin/python). I don't see such issues there.
Good information, thanks! I was thinking I had setup a really strange and unique configuration :-) I am wondering which version of glibc you have, or if you are using the LD_ASSUME_KERNEL variable in any way? Can you also tell me what your ./configure command line looked like when you compiled Python? I have compiled both 2.4 and 2.3.3 on the FC1 machine, but perhaps I left something critical out of the configuration.
Tres. - -- =============================================================== Tres Seaver tseaver@zope.com Zope Corporation "Zope Dealers" http://www.zope.com -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFCCLrIGqWXf00rNCgRAphPAKCOgJqHZzBZZH7xdYENY05nU3p/EwCfQ4WH ZqdpeRLmsx/Ugd4n31pePYU= =M1/1 -----END PGP SIGNATURE-----
Thanks again, David ---------------------------------------------------------------- This message was sent using IMP, the Internet Messaging Program.