incubator-couchdb-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Paul Davis <>
Subject Re: migrating data from a previous couch version
Date Sun, 24 May 2009 18:48:00 GMT
I'm not at all certain and I could see it going either way. But
assuming both architectures are Intel I would hazard a guess to say
that it'd be fine. If you Mac is PPC then I'd lean towards it not
working. But my endian-fu is not what it used to be. Also, this is
assuming that the Erlang binary format doesn't take care of endianess
and I haven't looked. It could be that everything is already stored in
network byte order and it wouldn't make a bit of difference.

Bottom line, I'd give it a shot.


On Sun, May 24, 2009 at 2:41 PM,  <> wrote:
> Is the format platform independent?  Can I do all this on muy mac altho the data was
orig created on ubuntu? Nitin
> Sent via BlackBerry by AT&T
> -----Original Message-----
> From: Paul Davis <>
> Date: Sun, 24 May 2009 13:08:19
> To: <>
> Subject: Re: migrating data from a previous couch version
> Nitin,
> I would just pull a version of the subversion repository a couple
> revisions before 753448. We didn't break the format twice or anything
> so the fact that it doesn't work probably means that you had one prior
> to that commit.
> I'm pretty certain that was the commit that introduced the versioning
> information so I can't point at anything better.
> Paul Davis
> On Sun, May 24, 2009 at 1:02 PM, Nitin Borwankar <> wrote:
>> Paul Davis wrote:
>>> Nitin,
>>> There's some info on how to upgrade on the wiki at [1] and [2]. If
>>> your 0.9aXXXXXX version has XXXXXX < 753448 you'll need to use the
>>> Ruby script at [2]. Otherwise you should be able to use couchdb-dump
>>> and couchdb-load from couchdb-python or run two couchdb nodes and
>>> replicate from the old version to the new version. If you're two
>>> versions span that commit then you must use the Ruby script because
>>> replication changes as well.
>>> Paul Davis
>> Hi Paul,
>> Thanks much for those links and I hope to get to the point where I can use
>> them.  Currenty I only have my data in files with
>> a) no older couch instance running and what's worse,
>> b) no idea what version I was running in my old couch, although I know it
>> was a 0.9a<something> from trunk.
>> so I guess my question becomes
>> a) how do I know just from the database binary files what exact couch
>> version created them   - is there some util or a specific set of bytes in
>> the header or
>> some voodoo that I can use and then I can go install that specific couch
>> version from src.  Then use the database dump/load scripts.
>> Nitin
>>> [1]
>>> [2]
>>> On Sun, May 24, 2009 at 5:13 AM, Nitin Borwankar <>
>>> wrote:
>>>> Hi guys,
>>>> This a rather strange situation but I need to get out of it :-)
>>>> I had a few Gig of data in a 0.9axxxxxx version.
>>>> I wanted to upgrade couch to 0.9 tarball when it was released.
>>>> I wasn't sure how it would go so I made a filesystem level backup of the
>>>> data directories.
>>>> Then when 0.9 installed and was running fine I copied one of the
>>>> databases
>>>> from the old backed up dir to the new 0.9 couch's data dir.
>>>> I got a "bad version" error from which I gather that the older version
>>>> data
>>>> is somehow not binary compatible with the newer one.
>>>> I don't have a running older version couch - and it was installed from
>>>> src
>>>> anyway and I have no idea what minor rev of 0.9a it was.
>>>> So how do I recover the old data from my files and upload it to the new
>>>> version.
>>>> Any ideas ?
>>>> There's a lot of data in there and I absolutely need to get it out -
>>>> help!
>>>> Nitin
>>>> 37% of all statistics are made up on the spot
>>>> -------------------------------------------------------------------------------------
>>>> Nitin Borwankar

View raw message