[ZODB-Dev] bug in ZEO-1.0b3

[Jeremy Hylton]

>>>>> "DB" == Dyon Balding <dyon@devcoder.com.au> writes:

  DB> If the ZEO server goes down, the the client no longer tries to
  DB> keep reconnecting.  This works fine with the ZEO-1.0b1 (I
  DB> haven't tried b2).

  DB> This appears to be something to do with the __closed attribute
  DB> in zrpc.py.  This is set to 1 when the server goes down, and the
  DB> client only tries to reconnect when it's value is 0.

  DB> Can anyone else verify this bug?

I can verify that there is at least one problem here, but I'm not sure
what its cause is.  

I assume you suspect the closed attribute in ClientStorage, as zrpc
doesn't have any attribute with close in the name and __closed is part
of the StorageServer.  I'll admit it as a possibility since it's new,
but it's only set when the ClientStorage's close() method is called,
which shouldn't happen when the server goes down.

If you've got the energy, could you capture a zLOG trace of this
failure to reconnect?  If there's a "INFO(0) ClientStorage close"
message in the log, then you're probably right.

  DB> Is there a page to check the status of bugs in ZEO? 

Good question.  I'm not aware of one, but perhaps Jim can point us at one.

  DB> When is the next version due out? 

It was due out tomorrow, but you had to go and find a bug <wink>.  The
next version should be ready as soon as all the bugs are fixed, and
this reconnection bug is the chief one.  There are a couple of other
bugs that have been reported on this list recently that also need to
be addressed.

  DB> Will it be a final 1.0 version?

If we're confident that the fixes for the current batch of bugs are
minor enough that enough beta isn't justified.  I hope it's a final
version. 

Jeremy