couchdb-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Nick Vatamaniuc <>
Subject [VOTE] Couch replicator _scheduler/jobs and _scheduler/docs API change
Date Thu, 31 Oct 2019 19:45:50 GMT
Hi everyone,

For CouchDB 3.0 I am proposing to add all the replication status
details from _active_tasks like "docs_written", "docs_read", etc to
_scheduler/jobs and _scheduler/docs endpoints.

Couch replicator scheduler has been out for a few years already. It
added flexibility, performance and other features. One of the goals of
the scheduler was to provide a better set of APIs for monitoring
replication job progress. Those APIs are _scheduler/jobs and
_scheduler/docs on top of the already existing _active_tasks and just
polling _replicator db documents.

_scheduler/jobs is focused on replication jobs, their history, how
many times they started, stopped, their state. It returns replications
jobs indexed by their job ids, and includes jobs started from
_replicator dbs as well as from the _replicate HTTP endpoint.

_scheduler/docs in turn, is focused on showing the state of
replication documents. It returns replication jobs indexed by
_replicator db and doc id which started the jobs.

_scheduler/docs API was returning an "info" field for each job. That
info field was designed to contain details pertaining to the current
state of the job. So if there is an error it will have the error
message. For a completed job, it was showing the replication stats
persisted to the replication document, so that users do not have to
poll their status.

However, when the jobs were running or were in a pending state that
field was "null". The idea was to eventually replace "null" with stats
from _active_tasks such that users don't need to ever look at
_active_tasks to monitor their replication progress. So the proposal
here is to do exactly that.

To be specific:

1) In the "pending" or "running" state _scheduler/docs will return
active_tasks status values in the "info" field. ("completed" is
already doing that btw).

2) _scheduler/jobs will have a new "info" field which also show the
same status values

There is a draft PR for it

Some examples of what it looks like:

What does everyone think?

Let's do a lazy consensus since we are replacing a place-holder null
value and adding a new field to _scheduler jobs. Let me know if this
is not appropriate and we need a different voting / discussion


View raw message