lucene-solr-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dmitry Kan <solrexp...@gmail.com>
Subject Re: redo log for solr
Date Fri, 29 Aug 2014 10:37:30 GMT
@Shawn: the suggestion about custom UP you made proves to be really useful.
Thanks!

It is also great to learn the internal solr API of searching documents
(need to find documents in Solr and store in a paper trail before they get
deleted).

Dmitry


On Thu, Aug 28, 2014 at 4:25 PM, Dmitry Kan <solrexpert@gmail.com> wrote:

> It may mean that I wasn't clear enough :)
>
> The idea is to build a paper trail system (without negative connotation!).
> Such that for instance if user deleted some data _by mistake_ and we have
> hard-committed to solr (upon which the tlog has been truncated), we paper
> trail'ed the document before the delete for providing the restore
> functionality.
>
> So if tlog is meant to make soft commits durable, this feature will serve
> more like undo functionality and persist the _history_ of modifications.
>
> I'm currently investigating what you suggested over IRC -- the
> UpdateProcessor. Looks like a way to go.
>
> Thanks,
>
> Dmitry
>
>
> On Thu, Aug 28, 2014 at 4:16 PM, Shawn Heisey <solr@elyograg.org> wrote:
>
>> On 8/28/2014 3:10 AM, Dmitry Kan wrote:
>> > We have a case when any actions a user did to the solr shard should be
>> > recorded for a possible later replay. This way we are looking at per
>> user
>> > replay feature such that if the user did something wrong accidentally or
>> > because of a system level bug, we could restore a previous state.
>> >
>> > Two actions are available:
>> >
>> > 1. INSERT new solr document
>> > 2. DELETE existing solr document
>> >
>> > If user wants to perform an update on the existing document, we first
>> > delete it and insert a new one with modified fields.
>> >
>> > Are there any existing components / solutions in the Solr universe that
>> > could help implement this?
>>
>> I'm wondering what functionality you need beyond what Solr already
>> provides ... because it sounds like Solr already does a lot of what you
>> are implementing.
>>
>> Solr already includes a transaction log that records all changes to the
>> index.  Each individual log is closed when you do a hard commit.  Enough
>> transaction logs are kept so that Solr can replay at least the last 100
>> transactions.  The entire transaction log is replayed when Solr is
>> restarted or a core is reloaded.
>>
>> What you describe where you delete an existing document before inserting
>> a new one ... Solr already has that functionality built in, using the
>> uniqueKey.  That capability is further extended by the Atomic Update
>> functionality.
>>
>> You're not new around here, so I don't think I'm telling you anything
>> you don't already know ... which may mean that I'm missing something. :)
>>
>> Thanks,
>> Shawn
>>
>>
>
>
> --
> Dmitry Kan
> Blog: http://dmitrykan.blogspot.com
> Twitter: http://twitter.com/dmitrykan
> SemanticAnalyzer: www.semanticanalyzer.info
>



-- 
Dmitry Kan
Blog: http://dmitrykan.blogspot.com
Twitter: http://twitter.com/dmitrykan
SemanticAnalyzer: www.semanticanalyzer.info

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message