A couple things.
- Shawn is correct, your location parameter should not include the bucket name. It's perfectly fine to have "s3://" included, but it should only contain the path within the bucket. - The other answers are correct that the directory must exist within the bucket for the backup to work, Solr will only create necessary directories inside if the base path already exists. - The Solr s3 backup module uses the default authentication/authorization credential provider chain: https://docs.aws.amazon.com/sdk-for-java/latest/developer-guide/credentials.html#credentials-chain The docs say that the "EC2 instance profile credentials" are used last, so this should work if you are running Solr on the EC2 instance, and have none of the other options in that list specified. - Houston On Fri, Jan 14, 2022 at 8:27 AM Michael Conrad <[email protected]> wrote: > Did you create *both* the bucket and the backup destination folder in > the bucket? > > On 1/14/22 04:10, Sergio García Maroto wrote: > > Hi. thanks for replies. > > > > 1) Yes I have tried both things. > > I created bucket manually and still same issue. > > > > 2) I added required jars and section to solr.xml > > > > I am running this from a linux server which doesn't have credentials > file. > > Instead has a role attached to the EC2 instance. > > It's supposed to work to validate against S3 if Solr used AWS SDK but I > > have the problem that's not available. > > > > Can anyone confirm that way of authentication is available? > > > > Thanks > > > > > > > > On Thu, 13 Jan 2022 at 17:47, Shawn Heisey<[email protected]> wrote: > > > >> On 1/13/22 9:09 AM, Sergio García Maroto wrote: > >>> I am trying to utilize new AWS S3 bucket backup contrib. > >>> Using the collection API I am getting the following error > >>> > >>> API CALL > >>> > >> > http://servername:8983/solr/admin/collections?action=BACKUP&name=personbackup&collection=person&repository=s3&location=s3://bucketname/backupfolder > >> > >> I recently did a test using the s3 backup repo, to help somebody figure > >> it out. I had it working. > >> > >> I believe that the "location" parameter should be a directory name > >> inside the bucket, not a URL. I do not recall whether I had to create > >> that directory in the bucket before it would work, or whether it created > >> the directory for me. If it were me, I would create the directory in > >> AWS prior to running the API call. > >> > >> I am assuming that you have taken steps to define the AWS access keys on > >> the startup command, and that you also took steps to add the required > jars? > >> > >> Thanks, > >> Shawn > >> > >> > >> >
