Hi all, Is zope.server still largely used by e.g. bluebream, grok, and other zope 3 apps? Or do people tend to use other WSGI servers instead? Either way, arguments for or against zope.server would be useful; I'm trying to decide whether to base some new stuff on it. - C
On Sat, Dec 17, 2011 at 09:57:18PM -0500, Chris McDonough wrote:
Is zope.server still largely used by e.g. bluebream, grok, and other zope 3 apps? Or do people tend to use other WSGI servers instead? Either way, arguments for or against zope.server would be useful; I'm trying to decide whether to base some new stuff on it.
We still use zope.server in production, mostly out of inertia (the app had been using Zope 3.2, then the 3.4 KGS for a looong time, before finally making the climb to ZTK 1.0). These are not high-volume sites. It surprises me occasionally with weird assertion errors of this kind: 2011-11-01T18:00:11 ERROR zope.server.taskthreads Exception during task Traceback (most recent call last): File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/taskthreads.py", line 49, in handlerThread task.service() File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/serverchannelbase.py", line 205, in service task.service() File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/http/httptask.py", line 69, in service self.channel.server.executeRequest(self) File "/apps/zopes/app-name/eggs/zc.zservertracelog-1.3.0-py2.6.egg/zc/zservertracelog/tracelog.py", line 145, in executeRequest task.write(response) File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/http/httptask.py", line 241, in write self.bytes_written += channel.write(data) File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/dualmodechannel.py", line 168, in write if not self._flush_some(): File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/dualmodechannel.py", line 185, in _flush_some num_sent = self.send(chunk) File "/usr/lib/python2.6/asyncore.py", line 358, in send self.handle_close() File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/dualmodechannel.py", line 59, in handle_close self.close() File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/dualmodechannel.py", line 206, in close assert self.async_mode AssertionError I haven't noticed any harm from this (the server keeps on serving other requests, clients do not see any error pages), but I'd like to know what it *means*. I thought I asked this list once, but received no reply -- but now I can't even find my question, so maybe I only planned to ask and never got around to it... This error happened once in 2010 and 24 more times in 2011. The other fun error I get (rather more often) is 2011-11-28T14:21:58 ERROR zope.server.taskthreads Exception during task Traceback (most recent call last): File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/taskthreads.py", line 49, in handlerThread task.service() File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/serverchannelbase.py", line 205, in service task.service() File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/http/httptask.py", line 77, in service self.channel.close_when_done() File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/dualmodechannel.py", line 193, in close_when_done while self._flush_some(): File "/apps/zopes/app-name/eggs/zope.server-3.6.2-py2.6.egg/zope/server/dualmodechannel.py", line 185, in _flush_some num_sent = self.send(chunk) File "/usr/lib/python2.6/asyncore.py", line 352, in send result = self.socket.send(data) error: [Errno 32] Broken pipe I suspect I know what this is (we do a redirect in self.update() but self.__call__() has no check for that and renders the template anyway, but the browser sees the redirect and closes the connection without reading the entire response body). I haven't actually proven that to my satisfaction, but, once again, there do not seem to be any harmful consequences (other than mildly annoying me with these rather rare error emails). I wish zope.server at least would let me know which request (method and URL) was responsible for the particular task (a __traceback_info__ annotation would make it into our error emails). Some day I might make that happen. Marius Gedminas -- http://pov.lt/ -- Zope 3/BlueBream consulting and development
* Chris McDonough <chrism@plope.com> [2011-12-18 03:57]:
Is zope.server still largely used by e.g. bluebream, grok, and other zope 3 apps? Or do people tend to use other WSGI servers instead?
We're using zope.server on several Zope 3 projects (that were begun before grok or WSGI or anythin existed ;), and it's *ahem* serving us just fine, no hiccups, no nothing, just doing its job. (We, too, see the occasional "broken pipe" error in our logs like Marius mentioned, I think that's due to requests that were aborted client-side, thus more an annoyance than anything.) On the other hand, we still haven't found a WSGI server we're happy with for deployment (as opposed to development). Paster has been known to randomly die on us (yup, I've only got anecdotes and nothing hard and fast, sorry), and the whole area of daemon management, logging, logrotation and so forth (not to mention buildout integration) seems not as advanced as we've come to know. What do people use here, what are your experiences and ideas? Wolfgang -- Wolfgang Schnerring · email/jabber: ws@gocept.com · software development gocept gmbh & co. kg · forsterstraße 29 · 06112 halle (saale) · germany http://gocept.com · tel +49 345 219401 0 · fax +49 345 1229889 1 Python, Pyramid, Plone, Zope - consulting, development, hosting, operations
On 20/12/2011 07:08, Wolfgang Schnerring wrote:
as advanced as we've come to know. What do people use here, what are your experiences and ideas?
What's wrong with mod_wsgi? Chris -- Simplistix - Content Management, Batch Processing & Python Consulting - http://www.simplistix.co.uk
On Tue, Dec 20, 2011 at 2:12 AM, Chris Withers <chris@simplistix.co.uk> wrote:
On 20/12/2011 07:08, Wolfgang Schnerring wrote:
as advanced as we've come to know. What do people use here, what are your experiences and ideas?
What's wrong with mod_wsgi?
Debugging with it seems to be a pain, so you end up using different servers in development and production. Jim -- Jim Fulton http://www.linkedin.com/in/jimfulton
On 20/12/2011 10:44, Jim Fulton wrote:
What's wrong with mod_wsgi?
Debugging with it seems to be a pain, so you end up using different servers in development and production.
I'm pretty sure if you ask Graham there'll be a simple way to run Apache & mod_wsgi in a way that lets you debug in much the same way as running ZServer or Paste. cheers, Chris -- Simplistix - Content Management, Batch Processing & Python Consulting - http://www.simplistix.co.uk
On Tue, Dec 20, 2011 at 08:08:36AM +0100, Wolfgang Schnerring wrote:
* Chris McDonough <chrism@plope.com> [2011-12-18 03:57]:
Is zope.server still largely used by e.g. bluebream, grok, and other zope 3 apps? Or do people tend to use other WSGI servers instead?
We're using zope.server on several Zope 3 projects (that were begun before grok or WSGI or anythin existed ;), and it's *ahem* serving us just fine, no hiccups, no nothing, just doing its job.
(We, too, see the occasional "broken pipe" error in our logs like Marius mentioned, I think that's due to requests that were aborted client-side, thus more an annoyance than anything.)
On the other hand, we still haven't found a WSGI server we're happy with for deployment (as opposed to development). Paster has been known to randomly die on us (yup, I've only got anecdotes and nothing hard and fast, sorry), and the whole area of daemon management, logging, logrotation and so forth (not to mention buildout integration) seems not as advanced as we've come to know. What do people use here, what are your experiences and ideas?
We've been using mod_wsgi for many moons. It's certainly very stable with piles of features for managing applications. Right now we're thinking about moving to Twisted because some bits of the application are asynchronous and already written on Twisted. It simplifies things to only have one daemon running everything. -- Brian Sutherland
On Sat, 2011-12-17 at 21:57 -0500, Chris McDonough wrote:
Hi all,
Is zope.server still largely used by e.g. bluebream, grok, and other zope 3 apps? Or do people tend to use other WSGI servers instead? Either way, arguments for or against zope.server would be useful; I'm trying to decide whether to base some new stuff on it.
FWIW, just to close this out, I poked and prodded zope.server into a fork named waitress (which only does WSGI). http://pypi.python.org/pypi/waitress The detailed differences between waitress and zope.server are here: http://docs.pylonsproject.org/projects/waitress/en/latest/differences.html I had originally thought I'd backport some of the changes to zope.server, but TBH the changes are pretty fundamental, and would require a ton of work to backport. Instead, if you like the features it provides (and can live with the ones it removes), I'd suggest using waitress instead of zope.server at this point. - C
participants (6)
-
Brian Sutherland -
Chris McDonough -
Chris Withers -
Jim Fulton -
Marius Gedminas -
Wolfgang Schnerring