[
https://issues.apache.org/jira/browse/SOLR-7583?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14568459#comment-14568459
]
Greg Solovyev edited comment on SOLR-7583 at 6/2/15 3:42 AM:
-------------------------------------------------------------
Hi [~andyetitmoves], I thought of adding an "output" kind of option to backup
command, but given that backup is asynchronous, it didn't make sense to block
the request and hold it until a backup completes. Maybe, the alternative to
current implementation is to stream directly back to the response without an
intermediate step of copying files and compressing them. The downside of
streaming the zipped archive directly back is that when a download fails due to
network interruption, it won't be possible to followup with a "continue"
request that will skip to the last byte in the previously requested backup and
continue streaming where previous request dropped off. With current approach,
because the backup is stored and is immutable, it is possible to add an option
to continue the download where it was left of and to add checksum verification.
was (Author: grishick):
Hi [~andyetitmoves], I thought of adding an "output" kind of option to backup
command, but given that backup is asynchronous, it didn't make sense to block
the request and hold it until a backup completes. Maybe, the alternative to
current implementation is to stream directly back to the response without an
intermediate step of copying files and compressing them. The downside of
streaming the zipped archive directly back is that when a download fails due to
network interruption, it won't be possible to quickly skip to the last byte in
the previously requested backup unless that backup is copied somewhere and
zipped before streaming starts. With current approach, it is possible to add an
option to continue the download where it was left of and to add checksum
verification.
> API to download snapshot files/restore via upload
> -------------------------------------------------
>
> Key: SOLR-7583
> URL: https://issues.apache.org/jira/browse/SOLR-7583
> Project: Solr
> Issue Type: Improvement
> Components: SolrCloud
> Reporter: Greg Solovyev
> Assignee: Mark Miller
> Attachments: SOLR-7583.patch, SOLR-7583.patch, SOLR-7583.patch
>
>
> What we are looking for:
> SolrCloud and Solr should have APIs to download a snapshot via HTTP.
> For single node Solr, this API will find a snapshot and stream it back over
> HTTP. For SolrCloud, this API will find a Replica that has the snapshot with
> requested name and stream the snapshot from that replica. Since there are
> multiple files inside a snapshot, the API should probably zip the snapshot
> folder before sending it back to the client.
> Why we need this:
> this will allow us to create and fetch fully contained archives of customer
> data where each backup archive will contain Solr index as well as other
> customer data (DB, metadata, files, etc).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
