Re: [Zope] One big Zope or many small Zopes? Or perhaps a ZEO?
Thanks Sean (and Kirk).
So, ya'll are saying we should have ONE zope managed over several servers with ZEO (and caching)? Key question is the ONE zope.
Currently, our huge zope is for the newspaper content, then homes is on it's own install, classifieds another and business directory another and two testing zopes on two others. (All on one box with our database pulled from another box.)
Am I understanding that correctly?
One other quick question. With 75,000 page views per day, set up the way we are now, (one box/multiple zopes) how many connections should we be allowing? I think the default it 4 and we have opened it up to 8. Sound right?
we're a newspaper house too, and currently running on one zope, with many zeo clients. what we're missing now, (what we should have done earlier, if we'd known :)) is to enable caching on dynamic sections, as suggested by a message before this. putting squid in front, as suggested by sean is also one of the answer. caching expensive methods is another. however, we're currently on 2.4.x, and RAM cache dont have automated invalidation. we're in the midst of trying to pull StandardCacheManager from 2.4.x to 2.3.x and hope that it doesnt break zope. so, if youre starting fresh, work with zope2.4.x, use ramcache, and probably use squid. with one zope, management should be easier. if the site crawls, add another zeo client. however, getting the logs together may be a challenge:) increasing the number of threads dont really guarantee faster response. there's a 7(10?) connection pool limit in the ZODB. in previous threads, we've tried increasing the thread number, and bumping up the connection pool to thread # + 2. http://zope.nipltd.com/public/lists/zope-archive.nsf/ByKey/09FAF907CB95FDF9 so if we're using -t 7, we increased the connection pool to 9. are you front ending with apache? from our experience, with > 100 simultaneous connections per zeo client, the site will crawl. we're experiencing that now, and thus the experiment with ram cache. *nix flavor recommended. sorry for the mish mash of information. it's the weekend :)
Our site is crawling too, hence the urgent messages, but thankfully, a new server is on it's way next week. Yes, apache is in front of zope 2.4.something now with the ramcache for various processes at 5, 10 and 20 minutes. What newspaper? Bill On 10/20/01 8:13 AM, "Bak @ kedai" <kedai@kedai.com.my> wrote:
Thanks Sean (and Kirk).
So, ya'll are saying we should have ONE zope managed over several servers with ZEO (and cach ing)? Key question is the ONE zope.
Currently, our huge zope is for the newspaper content, then homes is on it's own install, classifieds another and business directory another and two testing zopes on two others. (All on one box with our database pulled from another box.)
Am I understanding that correctly?
One other quick question. With 75,000 page views per day, set up the way we are now, (one box/multiple zopes) how many connections should we be allowing? I think the default it 4 and we have opened it up to 8. Sound right?
we're a newspaper house too, and currently running on one zope, with many zeo clients. what we're missing now, (what we should have done earlier, if we'd known :)) is to enable caching on dynamic sections, as suggested by a message before this. putting squid in front, as suggested by sean is also one of the answer. caching expensive methods is another. however, we're currently on 2.4.x, and RAM cache dont have automated invalidation. we're in the midst of trying to pull StandardCacheManager from 2.4.x to 2.3.x and hope that it doesnt break zope.
so, if youre startin g fresh, work with zope2.4.x, use ramcache, and probably use squid. with one zope, management should be easier. if the site crawls, add another zeo client. however, getting the logs together may be a challenge:)
increasing the number of threads dont really guarantee faster response. there's a 7(10 ?) connection pool limit in the ZODB. in previous threads, we've tried increasing the thread number, and bumping up the connection pool to thread # + 2.
http://zope.nipltd.com/public/lists/zope-archive.nsf/ByKey/09FAF907CB95FDF9
so if we're using -t 7, we increased the connection pool to 9.
are you front ending with apache? from our experience, with > 100 simultaneous connections per zeo client, the site will crawl. we're experiencing that now, and thus the experiment with ram cache.
*nix flavor recommended.
sorry for the mish mash of information. it's the weekend :)
participants (2)
-
Bak @ kedai -
Bill Blevins