nickva edited a comment on issue #1734: Stop replications on target document 
write failures
URL: https://github.com/apache/couchdb/pull/1734#issuecomment-438377284
 
 
   @rnewson 
   
   > there are many legitimate reasons for a document write to fail and for the 
replication to continue regardless. 
   
   The `continue regardless` thing is what this is mostly about, it configures 
whether continuing should happen or shouldn't happen. Once it continues and 
checkpoints it's a headache to retry it (delete checkpoints on target and 
source, create / update the replication document, get another error, do the 
same thing again, etc). A replication "successfully" completing is misleading 
the user as well in  a way and can be perceived as a "data loss".
   
   >  If the count of failed writes is not persisted, it should be. 
   
   We fixed that recently and now is persisted between job runs (before it 
wasn't) but persisting it to the replication doc would invite more updates to 
doc which were trying to move away from, especially with continuous 
replications. We could persist it to the the checkpoint history, we already 
save it there. But we'd have to checkpoint and retrying the replication would 
involve all the hassle as described above. We could invent some other, new 
place to persist replication state (new type of local replicator docs).
   
   I am not against adding more detailed counts, but that could be orthogonal 
to the idea of stopping a replication job on failure. The downside is that it 
doesn't help with knowing which document failed and why. So even then, fixing 
it means doing a diff between databases or scouring log files.

----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
 
For queries about this service, please contact Infrastructure at:
[email protected]


With regards,
Apache Git Services

Reply via email to