Thanks Shawn for your insights,
I had already considered the concerns shared by you, still couldn't find a
way to use _version_ while deleting or even querying a document.
Is there anything else running that might update the document between
the query request and the delete request?
No, there was no indexing running. I had even committed all the changes
before checking the version of documents. So, I am pretty sure there was no
chance the document was updated b/w the time I queried and deleted the
document.
Do you possibly have the same uniqueKey value in more than one shard?
No, we do not documents with same unique key on more than one shard. I
verified that the document I was trying to delete was on single shard (by
routing request on each shard one by one; *eg. _route_:shard1*; and
checking if document is present on any) before trying to delete the
document.
Just do any query that would match that document and the _version_ field
should be in the search results. If you have documents with the same
uniqueKey value in multiple shards, then which document gets returned
cannot be predicted. Solr will eliminate duplicates when returning results
so there is only one document for each uniqueKey field value.
I have extracted the _version_ field using query:
*fq=uniqueid:{some_value}&fl=_version_*, but upon running filter query
using this same version I extracted; fq=_version_:{extracted_version} it is
unable to find the document.
When using the implicit router, if you simply use deleteById and give it
the value in the uniqueKey field, I am reasonably certain that the delete
request is sent to all shards so it won't matter what shard it's in or what
_version_ it has.
Yes, we can delete documents without _version_, but I am still curious as
to why is delete request with _version_ is failing.
On Tue, Dec 19, 2023 at 10:05 PM Shawn Heisey <apa...@elyograg.org.invalid>
wrote:
> On 12/18/23 21:42, Saksham Gupta wrote:
> > I have written a small script in python to delete a product from solr
> cloud
> > collection on the basis of unique id, _route_ and _version_.
> >
> > I have extracted the exact values of unique id, _route_ and _version_
> from
> > solr index and used them to delete the product. But my script gives an
> > error which indicates that the _version_ I used is incorrect. The error
> > message looks like:
> >
> > "msg":"Async exception during distributed update: Error from server at
> > http://solr_ip:port/solr/collection_name/: null\n\n\n\nrequest:
> > http://solr_ip:port/solr/collection_name/\nRemote error message: version
> > conflict for {unique id} expected=1784635392440402000
> > actual=1784635392440401920",
> >
> > "code":409}}
> >
> > 1. How could this be the case where actual _version_ is different from
> the
> > one visible on querying.
>
> Is there anything else running that might update the document between
> the query request and the delete request?
>
> Do you possibly have the same uniqueKey value in more than one shard?
>
> If either of those conditions is true, then automatically generated
> fields like _version_ are subject to change.
>
> > 2. How to get the actual _version_ of a document?
>
> Just do any query that would match that document and the _version_ field
> should be in the search results. If you have documents with the same
> uniqueKey value in multiple shards, then which document gets returned
> cannot be predicted. Solr will eliminate duplicates when returning
> results so there is only one document for each uniqueKey field value.
>
> > 3. Any other ways to delete the solr document using _version_?
>
> When using the implicit router, if you simply use deleteById and give it
> the value in the uniqueKey field, I am reasonably certain that the
> delete request is sent to all shards so it won't matter what shard it's
> in or what _version_ it has.
>
> Thanks,
> Shawn
>
>
