On Tue, 06 May 2014, Petr Brož wrote:
> Looks like the virtual index cannot be directly reindexed but its
> components are reindexed together with the dependent indexes. But
> there appears to be some bug in the algorithm because the information
> in the virtual index can become stale after the reindexing of the
> dependant indexes as was in my case.

Here are some observations on this topic:

- `bibindex -u admin -w global -R` should trigger reindexing of the
  virtual global index.

- if you reach some stalled kind of situation, where the system things
  there is nothing to do, please try `--force` CLI parameter.

- If you have some virtual index, then it must be run as a separate
  bibindex process.  This was done for multi-node multi-process setup:
  e.g. you could have independent indexes running in parallel in some
  nodes, and all "pushing" to the virtual global index queue new terms,
  so the global indexing process would simply collect them from the
  queue.

  E.g. with Invenio defaults, two bibindex processes are running: one
  without arguments for the set of regular indexes (author, title, etc),
  and one virtual global index (`-w global`).

- There was indeed a period in time when the "next" branch was missing
  the second process; see for example:

  
https://github.com/inveniosoftware/invenio/commit/fdea8295f0a26a48fb1976e64a3fbbe6297dcf0b

Best regards
-- 
Tibor Simko

Reply via email to