Skip to content
Snippets Groups Projects
Commit 147ef843 authored by Eric Cano's avatar Eric Cano
Browse files

Update doc/migration/migrationFromCASTOR.md

parent cc37be35
No related branches found
No related tags found
No related merge requests found
Pipeline #13786 failed
......@@ -75,11 +75,13 @@ Any?
1. Stager shutdown means no read, no write, but move and delete are still possible (nameserver operations).
Other scenario:
0. User changes procedures and stops relying on the CASTOR disk cache.
1. Remove write rights for CASTOR
1. B/W lists in stager
2. archive route removal
3. Tapes marked readonly (FULL|RDONLY) (option new state/column in the DB)
4. Expriment commit to not change the files (we rely on this)
5. Experiments are aware that the disk cache will be lost in case of rollback.
2. ND import for CASTOR to CTA
1. import to EOS NS in parallel to next
2. import to CTA catalogue
......@@ -92,6 +94,26 @@ Other scenario:
4. Look in CASTOR NS logs for renames/delete (and apply to CTA?)
5. Scan CASTOR NS for differences
Rollback: CASTOR comes back in the state it was in at the beginning of the migration.
Redirection: It opens many scenarios. Is it avoidable? If fast enough, we could propose a downtime to experiments. => We need to assess achievable speed for the migration (and if fast enough favor a downtime)
Namespace import:
1. We need to create directories first, and then import files
2. Directory mapping strategy: this should take into account the existing pathes in the expriment's catalogues. (/.../castor/... symlink to the new tree could be a solution).
3. Problem of possible collision between new files (pure CTA) and imported CASTOR files.
1. We can have a special new directory, and merge after import.
2. Collisions have to be sorted out with experiments.
4. Oracle to Oracle could be achieved on the same DB server to optimize speed.
1. DB link?
2. DB bulk C/Python interface
5. EOS NS import streaming possible according to Andreas.
6. ACLs import: do we convert? drop? recreate based on expriemts rules.
7. Import via a protobuf stream, channel to be decided, will be done with Georgios.
Files IDs will be preserved: ArchiveFileId of imported files will be the CASTOR file Id.
The maximum file Id is 17 decimal characters, or roughly 50-somehting bits, so we can afford to reserve a range of few billions for imported files (1B ~ 30bits)
## Other requirements
Monitoring of files on tape should be created in continuity with CASTOR's.
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment