incubator-couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "R.J. Steinert" ...@rjsteinert.com>
Subject Re: How to configure a limit on simultaneous replications?
Date Fri, 03 Jan 2014 16:11:01 GMT
We're also seeing this issue on computers with more power than the
Raspberry Pi. The following `/replicator/_all_docs?include_docs=true`,
`_config`, and `_active_tasks` are all from an ASUS R503U Laptop with
the following specs:

AMD E2-1800, 1.7GHz, 2GB RAM, 320GB Hard drive

CPU Load is at 50%, Memory Usage is 50%.  The `werl (32 bit)` process
account for 1.4% CPU load and 30MB of Memory Usage.

As you can see in active_tasks.json, there are about 17 active
replication tasks. Some have varying success but the status of the
replications do not change for days on end. This laptop is connected
to an extremely fast connection (Cambridge Innovation Center next to
MIT) connecting with the Cloudant servers. After reducing the
`49f8c6cf5587c1504c54121997ca1ff5` doc, the replication processes just
fine.

- R.J. Steinert (https://github.com/rjsteinert)


On Mon, Dec 16, 2013 at 11:13 AM, R.J. Steinert <rj@rjsteinert.com> wrote:
> @Jens We're testing over a strong Internet connection. My laptop for example
> has no problem doing this replication in the couple of seconds you have also
> experienced, the Raspberry Pi on the other hand...
>
> @Robert Thanks for chiming in. Unfortunately I'm no Erlang expert so I won't
> be much help. I may however look at writing a replication manager in
> Javascript, perhaps contributing it to the PouchDB project.
>
>
> On Tue, Nov 26, 2013 at 3:33 PM, Robert Newson <rnewson@apache.org> wrote:
>>
>> If you mean the _replicator database then, yes, currently it will
>> attempt to run all the jobs at once. That is, there is Considerable
>> Room For Improvement.
>>
>> B.
>>
>>
>> On 26 November 2013 19:10, Jens Alfke <jens@couchbase.com> wrote:
>> >
>> > On Nov 26, 2013, at 10:07 AM, R.J. Steinert <rj@rjsteinert.com> wrote:
>> >
>> >> when CouchDB tries to replicate them all at once, the replications
>> >> never seem to finish in a dependable fashion. Sometimes some of the
>> >> databases will finish replicating others will hang for days. Each
>> >> database
>> >> is only a couple dozen KB so these replication processes aren't
>> >> managing a
>> >> whole lot of data.
>> >
>> > That’s weird. What sort of network connection is the replication using?
>> > Is this over a LAN (WiFi or Ethernet), a wider-range Internet connection, a
>> > cell network…? Basically the slower / laggier / flakier the network is, the
>> > more roadblocks are thrown in the way of the replicator. It should still
>> > work reliably, though, so what you’re describing sounds like a bug; a dozen
>> > tiny databases should be able to replicate in a few seconds unless there are
>> > major network issues.
>> >
>> > (Disclaimer: I am not familiar with the replicator implementation in
>> > CouchDB itself, although I have implemented compatible replicators so I know
>> > the protocol well.)
>> >
>> > —Jens
>
>

Mime
View raw message