Change Details

The software, home-made automation or third party solution (or a combination of both), should be ready to be deployed on WMF infrastructure and be able to backups 1 file and recover into Swift (or a test Swift installation) that same file. To the point, that it should quickly be able to backup and recover a small wiki such as testwiki. This will inform us of the following open questions: * Throughput needed/achieved and parallelization for both backup and recovery * Actual API used (Swift vs. MediaWiki) * Querying cache layer for speedup? * Desired recovery methodology (recover with the same name or different name/location?) * Overhead on Swift infrastructure/need for expansion to accommodate backup needs There won’t be any specialized storage for this proof of concept- as it will be small scale, we can just use a directory on a filesystem for temporary storage

As a person responsible for the WMF backups I want to test the dependencies of the planned media storage backup solution. This can be achieved through a proof of concept implementation that will interact with the existing infrastructure. Acceptance criteria: - A backup solution capable of backing up and restoring a single file from a specific wiki (e.g. testwiki) between Swift and a local filesystem is deployed to a test host - Testing covering the following areas is performed: - Throughput needed/achieved and parallelization for both backup and recovery - Which API to use (Swift vs. MediaWiki) - Querying cache layer for speedup? - Desired recovery methodology (recover with the same name or different name/location?) - Overhead on Swift infrastructure/need for expansion to accommodate backup needs

The software, home-made automation or third party solution (or a combination of both), should be ready to be deployed on WMF infrastructure and be able to backups 1 file and recover into Swift (or a test Swift installation) that same file. To the point,As a person responsible for the WMF backups I want to test the dependencies of the planned media storage backup solution. that it should quickly be able to backup and recover a small wiki such as testwikiThis can be achieved through a proof of concept implementation that will interact with the existing infrastructure. This will inform us of the following open questions: *Acceptance criteria: - A backup solution capable of backing up and restoring a single file from a specific wiki (e.g. testwiki) between Swift and a local filesystem is deployed to a test host - Testing covering the following areas is performed: - Throughput needed/achieved and parallelization for both backup and recovery * Actual - Which API to used (Swift vs. MediaWiki) * - Querying cache layer for speedup? * - Desired recovery methodology (recover with the same name or different name/location?) * - Overhead on Swift infrastructure/need for expansion to accommodate backup needs There won’t be any specialized storage for this proof of concept- as it will be small scale, we can just use a directory on a filesystem for temporary storage