Under clusterwide-file-manifest we should have directory tree of the form
dir - K-Y-Z-M:
K-Y-Z-M/clusterwide-manifest.nmd5
These are the clusterwide manifests - i.e manifests that are produced by comparing the md5s of different nodes from the same cluster, and producing 2 lists for each of those clusters:
-
List 1 (matches): md5sum:/file/path
-
List 2 (diffs): md5sum:/file/path
Where List 1 is a list of all files that have the same path, and have the same path and md5sum across the cluster While list 2 is a list of the inverse of the above.
Stages
Stage 0 - ssh backbone for propogating stage1 across the cluster
git clone https://git.adspem.com/opthq/nmd-md5sum-manifest.git
cd nmd-md5sum-manifest/stages
chmod +x *
./stage0-remote-run.sh
hsctl cmd run 'ls -la /root/manifest*'
Stage 1 - Md5:filepath
- Param 1 - base path (i.e
stage1.sh /- would produce md5sum for everything under root) - Produces a file containing the md5:file on a given system
- Excludes directories known to contain node-specific directories
Stage 2 - Base manifest producer -
Takes in files produced from Stage 1 across a cluster and produces an output base-manifest