incubator-couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Filipe David Manana <fdman...@apache.org>
Subject Re: Replication and checkpoints - what to expect?
Date Thu, 21 Jun 2012 19:40:16 GMT
On Thu, Jun 21, 2012 at 8:28 PM, Andreas Kemkes <a5sk4s@yahoo.com> wrote:
> The bump was by mistake, I wanted to add to it (i.e., stalling replication - see other
reply).  Thanks for answering.
>
> It should report 100% in my opinion if it has caught up with the current source sequence,
even though in the future there might be more.

That is the case for continuous replications. When target has caught
up with the source, it goes to 100%. When source gets new stuff,
target will report a progress < 100% for some time, and then
eventually back to 100%.

>
> The same should be true for filtered replications if there is no applicable document
between the current source sequence and the last checkpoint.  Otherwise you would be always
wondering if it has been replicated entirely.

That's harder. With filtered replication, we only know about sequence
numbers of changes that pass the filter.

>
> -- Andreas
>
>
> ________________________________
>  From: Robert Newson <rnewson@apache.org>
> To: user@couchdb.apache.org; Andreas Kemkes <a5sk4s@yahoo.com>
> Sent: Thursday, June 21, 2012 10:40 AM
> Subject: Re: Replication and checkpoints - what to expect?
>
>
> If you're running continuous replication, then the process isn't going to reach 100%
(since that would imply all future changes have been replicated, which is clearly impossible).
I suspect it's just a reporting anomaly. I didn't reply earlier because I hadn't had time
to verify the code but since you bumped the thread, I thought I'd best say something.
>
> B.
>
>
> On 21 Jun 2012, at 17:50, Andreas Kemkes wrote:
>
>> Reply to user@couchdb.apache.org
>>
>>
>> ________________________________
>> From: Andreas Kemkes <a5sk4s@yahoo.com>
>> To: "user@couchdb.apache.org" <user@couchdb.apache.org>
>> Sent: Tuesday, June 19, 2012 5:36 PM
>> Subject: Replication and checkpoints - what to expect?
>>
>> Upgrading to 1.2 on the target system allowed me to setup an unfiltered continuous
replication between a couchdb 1.1.1 instance and a couchdb 1.2 and then use the latter as
the source for several filtered continuous replications to shard a monolithic couchdb database.
>>
>> The unfiltered replication was started from the futon replicator api, all the others
from the command line using the ../_replicate call.  All replications are pull.
>>
>> The all work nicely, except that the replication between couchdb 1.1.1 and couchdb
1.2 seems to not write any checkpoints after a while:
>>
>> Checkpointed source sequence 911275, current source sequence 918089, progress 99%
>>
>> Is this to be expected?  Would an upgrade to couchdb 1.2 on the source system correct
the issue?
>>
>>
>> I also think that the filtered replications, which don't involve couchdb 1.1.1, only
checkpoint to the latest sequence that involves a document that passes the filter.  Even
if new documents are added at the source and replicated to the target, a new checkpoint is
not always written.
>>
>> Is this a known issue?
>>
>> How can I further narrow down the issue?
>>
>> Any documentation on the interplay of replication and checkpoints would helpful as
well.  Thanks for your help.
>>
>> -- Andreas



-- 
Filipe David Manana,

"Reasonable men adapt themselves to the world.
 Unreasonable men adapt the world to themselves.
 That's why all progress depends on unreasonable men."

Mime
View raw message