On 11/08/2020 13:21, Tim Flicker wrote:
On 8/11/2020 7:06 AM, Andy Seaborne wrote:
On 10/08/2020 22:33, Tim Flicker wrote:
Hi,
We are working on a backup / restore plan for TDB 1.0. It seems to be
working for xTDB. We recently upgraded to Jena 3.15.0 from 3.7.0 and
gTDB restore no longer seems to be working. We are backing up
connector files and doing a database dump on those collections to
TriG. During restore, we have a script that inspects the backup and
expands any TriG files back to a database folder / file structure
ending in .data using the tdbloader script and puts the connector
files back in place. The gTDB collections don't seem to be available
however. Are there other databases or files that need to be backed up
and restored in 3.15.0 that weren't needed in 3.7.0 for gTDB?
Regards,
Tim
Context:
gTDB (the TQ term for using TDB1 with the data in the default graph -
one DB, one graph).
xTDB (TQ term) is one database for multiple graph per database,
Tim - When you restore, what is in the database? Has the file sizes on
disk increased? What did tdbloader say? and tdbdump after reload? The
graph data does not go into a named graph, but into the default graph.
The files within the .data directory seem to be the same size as the
original project except for journal.jrnl, which is empty in the restored
project.
The connector files from the original and restored projects are identical.
tdbloader sample output on restore:
07:43:44 INFO loader :: -- Start triples data phase
07:43:44 INFO loader :: ** Load empty triples table
07:43:44 INFO loader :: -- Start quads data phase
07:43:44 INFO loader :: ** Load empty quads table
07:43:44 INFO loader :: Load:
./samples.teamwork.topbraidlive.org/country-codes/country-codes.tdb.data.trig
-- 2020/08/11 07:43:44 EDT
07:43:45 INFO loader :: -- Finish triples data phase
07:43:45 INFO loader :: ** Data: 5,960 triples loaded in 0.52
seconds [Rate: 11,439.54 per second]
07:43:45 INFO loader :: -- Finish quads data phase
07:43:45 INFO loader :: -- Start triples index phase
07:43:45 INFO loader :: ** Index SPO->POS: 5,960 slots indexed in
0.03 seconds [Rate: 192,258.06 per second]
07:43:45 INFO loader :: ** Index SPO->OSP: 5,960 slots indexed in
0.03 seconds [Rate: 192,258.06 per second]
07:43:45 INFO loader :: -- Finish triples index phase
07:43:45 INFO loader :: ** 5,960 triples indexed in 0.07 seconds
[Rate: 85,142.86 per second]
07:43:45 INFO loader :: -- Finish triples load
07:43:45 INFO loader :: ** Completed: 5,960 triples loaded in 0.60
seconds [Rate: 10,016.81 per second]
07:43:45 INFO loader :: -- Finish quads load
So that has loaded into the default graph.
No other program running I hope.
tdbdump after reload has no output
The loader log above shows that there are 5,960 triples going somewhere.
If tdbdump is empty, then it could because it is run on a different disk
location.
gTDB databases go in a different place to xTDB databases.
but it does create a database
structure with most files around 8KB.
Check the "nodes.dat" - it should be non-zero length.
Andy
(This does not look like a Jena TDB issue. )
Check the connector files are "unusual" - they should not name a graph
for gTDB.
Andy
