[ZCM] [ZC] 1521/ 3 Comment ""ZEO.Exceptions.ClientDisconnected" should retry"

[Collector: Zope Bugs, Features, and Patches ...]

Issue #1521 Update (Comment) ""ZEO.Exceptions.ClientDisconnected" should retry"
 Status Pending, Database/feature medium
To followup, visit:
  http://zope.org/Collectors/Zope/1521

==============================================================
= Comment - Entry #3 by mcdonc on Oct 1, 2004 9:21 am

Maybe instead just provide a hook to register an additional exception type and a retry policy for that exception type into a ZPublisher "retry registry".  When the ZPublisher exception hook encounters one of these exceptions, it performs a retry based on the retry policy for that exception type.  The policy could include the amount of time to sleep before actually doing the retry, the maximum number of retries, and so forth (there's another collector issue in here that wishes we could have a number-of-retries policy for conflicts; this would sort of be an extension to that).
________________________________________
= Comment - Entry #2 by htrd on Oct 1, 2004 8:52 am

Sounds like a good idea.

Do you have any experience of retrying ClientDisconnected errors in ZPublisher? I would think that it might take a few seconds for the ZEO client to reconnect, and an immediate retry would likely fail. 



________________________________________
= Request - Entry #1 by d.maurer on Oct 1, 2004 8:36 am

When ZEO restarts, its clients get a "ClientDisconnected" exception. It is usually delivered to the Zope end users.

As "ClientDisconnected" is a transient problem, the ZPublisher should probably treat this the same as "ConflictError".

I propose to define a new exception "TransientError" and
base "ZPublisher"s retrials on this exception.

Let "ConflictError" and "ClientDisconnected" (and maybe
other transient errors) be inheriting from this exception
==============================================================