O3/CNAF/3-registration.md

@@ -9,8 +9,16 @@ Two ways to register data:
  	 register all frames in those times
  * offline: select a start time and datatype, register frames as they arrive.
 
+There are already a couple of weeks of frames that we want to transfer, so
+we'll need to use the offline mode to register the already existing files, and
+the online mode to stay synchronised as frames arrive.
+
 For this exercise, I will register data directly at CIT and transfer to CNAF.
 
+The `gwrucio_registrar` tool, which we'll use to register frames in rucio,
+supports registration from lists of files and reading from a diskcache.  The
+diskcache mode is required for online operation (continuous registration), so
+let's use that for offline registration, too.
 
 ## DiskCache
 A diskcache daemon usually runs locally.  I'm not sure where that lives (or if
@@ -57,8 +65,42 @@ After a short time, there is an ascii-dump of the frame cache at:
 /home/jclark/Projects/rucio-O3/CNAF/diskcache/frame_cache_dump
 ```
 
-
 ## Offline registration
+First, create a **registration file**:
+```
+L-L1_HOFT_C00:
+  scope: "ER13"
+  regexp: "L-L1_HOFT_C00"
+  minimum-gps: 1228836864
+  maximum-gps: 1228902400
+  rse: LIGO-CIT-ARCHIVE
+```
+This tells `gwrucio_registrar` to:
+ 1. create a dataset called `L-L1_HOFT_C00` in the `ER13` scope
+ 1. register frames of type `L-L1_HOFT_C00` in the time range `[1228836864, 1228902400]`
+ 1. register files as existing at the `LIGO-CIT-ARCHIVE` RSE
+                                                               
+
+Run `gwrucio_registrar:`
+```
+singularity exec \
+              --bind /archive \
+              ../gwrucio-latest.simg \gwrucio_registrar \
+              -r ER13-L1_HOFT_C00.yml \
+              --verbose\
+              daemon --run-once \
+              /home/jclark/Projects/rucio-O3/CNAF/diskcache/frame_cache_dump
+```
+(Scripted [here](https://git.ligo.org/james-clark/gwrucio/blob/master/bin/gwruciod-once))
+
+Some notes:
+ * We run from a singularity container ultimately created from [this dockerfile](https://git.ligo.org/james-clark/gwrucio/blob/master/Dockerfile)
+ * the `--run-once` option in `daemon` mode means the code will exit after
+   registering all of the files found from the diskcache in `frame_cache_dump`
+   which match the criteria described in the reg-file `ER13-L1_HOFT_C00.yml`
+ * If `--run-once` is *not* provided, the code will re-read the diskcache every
+   30 seconds (by default) and register any new, un-registered files matching
+   the reg-file criteria.
 
 
 ## Online registration