[Bucardo-general] If a db is unreachable then nothing works!

[Greg Sabino Mullane]

So I want to make sure I understand the proposed solution. Everyone 
please feel free to chime in here. Teh current situation when 
a database goes down is:

Any KID accessing it dies. In theory, the CTL would resurrect this KID, 
but it won't because:

The MCP will also notice the dead database, and exit.

The proposed solution (hybrid of myself and Rainer's):

The MCP still watches over the databases, but merely makes a note 
when a DB is unreachable. If it determines that there is no valid 
source->target path left, it will not resurrect the CTL in question 
for each sync.

The CTL also notices failed databases, and continues on if there is 
still a valid source->target path. Otherwise, it simply exits. Otherwise, 
it will only keep watching over/spawning a subset of the KIDs.

Both the MCP and the CTL will periodically poll the failed databases, 
and put things back into rotation when they are back up.

That's my quick thoughts on the matter for B4. For B5, there would be
something similar, except that we no longer have one KID per path anymore, 
but one KID per sync, so that KID would have some brains to figure out 
what is going on and what targets to skip. And, of course, it would only 
skip targets, not sources.

-- 
Greg Sabino Mullane greg at endpoint.com
End Point Corporation
PGP Key: 0x14964AC8
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 163 bytes
Desc: not available
URL: <https://mail.endcrypt.com/pipermail/bucardo-general/attachments/20120429/686aec97/attachment.sig>