Github user nickva commented on the pull request:
https://github.com/apache/couchdb-couch-replicator/pull/29#issuecomment-192353044
@kxepal
`http_connections` is a bit misleading as it configures 3 separate things
-- httpc connection pool for source, for target and also constrains the number
of replication workers. Docs say configuration means "Maximum number of HTTP
connections per replication" but user ends up with two of them -- one for
target and one for source, even if they set it to 1
> So, either we need to forbid set http_connections to 1 ...
Because this configuration is overloaded, 1 is still valid for target http
connection pool and to constrain the number of replication workers. Here we
only adjust the size of the http connection pool for an http source. This is
also logged so if they care user can find out why they have an extra
connection.
> why 1 connection is not enough since it was fine for 1.x. Why replication
needs at least two of them?
I don't see `get_pending_changes` in 1.6.x branch that might explain why:
https://github.com/apache/couchdb-couch-replicator/blob/master/src/couch_replicator.erl#L308
vs
https://github.com/apache/couchdb/blob/1.6.x/src/couch_replicator/src/couch_replicator.erl#L280
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
