Conversation
4ee8872 to
a54ef34
Compare
6f91842 to
d6c8495
Compare
4f561ce to
cf2147e
Compare
3f5f323 to
e1eef38
Compare
e7ac1c3 to
bff4458
Compare
7ce5ab9 to
ce3299e
Compare
23eb8ee to
0480f8a
Compare
6fb9cca to
10a4deb
Compare
c80414f to
c963e3b
Compare
158db82 to
1759378
Compare
43b7244 to
810e4e4
Compare
|
I found a conceptional Problem, which leads to errors in the tiling process. The MCRTreeCopier commits and restarts database transactions. The tiling queue runs in another thread is now able to find the commited tiling jobs and now starts tiling. The ocfl transaction of the Thread in with the MCRTreeCopier did not yet commit its OCFL Transaction, so the files which should be tiled do not exist yet. Meanwhile the uploads happen in another transaction and are stored in a local directory, so the "long" transaction only runs long to upload from the mycore machine to s3, or to copy from one local to another local directory, which should not take that long. |
360ce0b to
f7d26f4
Compare
7eac844 to
947008b
Compare
2603b9e to
f0fd713
Compare
* add restore derivate command * add describe object command * fix MCRVersionedPath#getParent does not return a versioned path
- simplify virtual objects - rewrote remote storage to use digest instead of path - update junit test - improve javadoc
- cp from transactional storage to remote storage after commit - use "real" CAS files in remote storage
the first commit to a derivate was not stored on the remote storage due to wrongful calculation of the version path
in remote repos stream directly to target path when makeing a copy/externalCopy instead of streaming the source first and then copy from there
- makes the journal smaller - compact every hour to store LRU queue - memory is enough to store "touch" position
24ca4cf to
221d99c
Compare
Link to jira.