Chris Withers wrote:
Darian V Schramm wrote:
I'm getting many of these in my event.log
2005-03-15T10:17:56 INFO(0) ZEC:1-None-1 flipping cache files. new current = 1
This is nothing to worry abotu from a crashing point of view. Worry abotu it when you want to improve performance once you have a happilly running server.
Let's say that I have a happily running server, that I get quite a lot of "flipping cache files" with a cache-size at 6000 in my Zope instances, and that I want to improve performances. What should I do? Guillaume Tamboise
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Guillaume Tamboise wrote:
Chris Withers wrote:
Darian V Schramm wrote:
I'm getting many of these in my event.log
2005-03-15T10:17:56 INFO(0) ZEC:1-None-1 flipping cache files. new current = 1
This is nothing to worry abotu from a crashing point of view. Worry abotu it when you want to improve performance once you have a happilly running server.
Let's say that I have a happily running server, that I get quite a lot of "flipping cache files" with a cache-size at 6000 in my Zope instances, and that I want to improve performances. What should I do?
Bump up the size of your ZEO client cache until the flips are only infrequent. Note that the value for the optoin is a 'byte-size', with a default of '20MB'; 6000 is not likely an improvement over the default. ;) Tres. - -- =============================================================== Tres Seaver tseaver@zope.com Zope Corporation "Zope Dealers" http://www.zope.com -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.5 (GNU/Linux) Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org iD8DBQFCWSGUGqWXf00rNCgRAl0iAJ9C3OeVpXTZcLU3T5itqh+HHXwW8gCfRZQJ eVSd4iDgneJ2ZzaXO4e1T6E= =Yzc2 -----END PGP SIGNATURE-----
[Guillaume Tamboise]
Let's say that I have a happily running server, that I get quite a lot of "flipping cache files" with a cache-size at 6000 in my Zope instances, and that I want to improve performances. What should I do?
I suspect you're confusing the ZODB cache (an in-memory cache measured in number of objects -- 6000 is sane for that, although on the small side) with the ZEO client cache (a disk-based cache measured in bytes -- 6000 is way too small for that). The latter cache is the one that "flips". There isn't a "one size fits all" answer for either. Change, experiment, measure, repeat, using your data and your object access patterns on your boxes and your network. Bigger is better until it starts to hurt; the point at which it starts to hurt depends on endless details of your access patterns, OS, HW, etc.
participants (3)
-
Guillaume Tamboise -
Tim Peters -
Tres Seaver