Chris Withers wrote at 2005-7-15 20:38 +0100:
> ...
>If restarting a Zope client running of a storage server takes more than 
>one or two seconds on "normal" hardware, I reckon we should focus on 
>speeding that up rather than battling with the refresh stuff...

I can tell you that this is quite difficult:

  On my (3 years old) hardware, a restart costs about 5 s.

But, there are further reasons:

  *  colleagues of mine use a single Zope installation as
     a development platform for a group of developers.

     They do not want continued restarts to make product
     modifications visible.

     And they can not use Chris M's "bin/zopectl fg" approach,
     because only a single one could have the Zope process
     on its console (apart from the fact, that "zopectl fg"
     is inferior to detect errors because problem messages
     must be filtered out from other unrelated log messages).

     Refresh fits this bill very nicely...

  *  we use refresh even in a production environment:

     refresh is much less intrusive (as seen by our visitors)
     than a restart. Not only, it saves about 15s of
     restart time (restart is slower there because RAM is tight
     and the code needs to be loaded via NFS); it also
     keeps the session info intact (maintained in a RAM storage).

     As you see, we have high confidence in "refresh" :-)

For more information about ZODB, see the ZODB Wiki:

ZODB-Dev mailing list  -

Reply via email to