- Jun 12, 2017
-
-
Nicolas Dandrimont authored
-
- Jun 07, 2017
-
-
Antoine R. Dumont authored
-
Antoine R. Dumont authored
Before this commit, the following situation could arise: - no encoding detected (we cannot do much after that yet) - no result after multiple tryouts to compute the lang This resulted in exceptions at runtime. So, for now, we check those situations and we reference the sha1 with lang None. The purpose is to permit an ulterior scheduling of those later on (e.g. when we improve the encoding policy reuse for example).
-
- Jun 06, 2017
-
-
Antoine R. Dumont authored
When chunking on the wrong byte, we can raise 'unexpected end of data'. This permits to capture this error and test the same content on the next ending byte. Related T722
-
Antoine R. Dumont authored
-
Antoine R. Dumont authored
Related T721
-
Antoine R. Dumont authored
-
- Jun 02, 2017
-
-
Antoine R. Dumont authored
Related T722
-
Antoine R. Dumont authored
-
Antoine R. Dumont authored
Related T722
-
Antoine R. Dumont authored
-
Antoine R. Dumont authored
Related T722
-
- Jun 01, 2017
-
-
Antoine R. Dumont authored
It's up to the indexer to query the storage to determine the indexer configuration's identifier and pass along the information when sending data for filter or update operations. Related T722
-
- May 29, 2017
-
-
Antoine R. Dumont authored
-
Antoine R. Dumont authored
Related T722
-
- May 22, 2017
-
-
Antoine R. Dumont authored
-
- May 17, 2017
-
-
Antoine R. Dumont authored
Related T713
-
Antoine R. Dumont authored
-
Antoine R. Dumont authored
-
Antoine R. Dumont authored
No rescheduling by default
-
Antoine R. Dumont authored
Prior to this, each indexer received the same batch of contents as the orchestrator. As some of our indexer need quite some time to finish when the batch size is huge enough, the only reasonable step was to reduce such batch size. But, in effect, this impacted as well other indexers which did not need such restriction (this for example creates more little db transations). The introduction of such option avoids that impact. This also introduces the option to check the presence on a per indexer basis. Related T713
-
- May 16, 2017
-
-
Nicolas Dandrimont authored
-
- May 12, 2017
-
-
Antoine R. Dumont authored
Default to no rescheduling.
-
Antoine R. Dumont authored
-
- May 04, 2017
-
-
Antoine R. Dumont authored
Related T712
-
- May 03, 2017
-
-
Antoine R. Dumont authored
-
Antoine R. Dumont authored
-
- May 02, 2017
-
-
Antoine R. Dumont authored
Related T713
-
Antoine R. Dumont authored
Related T713
-
Antoine R. Dumont authored
Related T712
-
- Apr 28, 2017
-
-
Antoine R. Dumont authored
Related T712
-
Antoine R. Dumont authored
Related T712
-
Antoine R. Dumont authored
If there are no hashes to recompute, do not read the content from objstorage. This optimization will only be useful for the case of the parameter recompute_checksum being false (which is the case in production for now) Related T712
-
- Apr 27, 2017
-
-
Antoine R. Dumont authored
Related T712
-
- Apr 26, 2017
-
-
Antoine R. Dumont authored
-
Antoine R. Dumont authored
Related T712
-
Antoine R. Dumont authored
-
Antoine R. Dumont authored
-
- Mar 21, 2017
-
-
Antoine R. Dumont authored
Related T692
-
Antoine R. Dumont authored
-