When its first tries to write to the search index the index does not yet exist and this error is thrown, handled, and logged, then the index is created. As long as you are not seeing this repeated over and over in the log its not a problem.
When content is created or edited, the indexing of that content is as follows, records are queued into the mp_IndexingQueue table with all the needed info to write to the index including the file system path to the index. Then a task is fired off on a background thread so as not to block the ui and this backround thread processes the rows in the indexing queue by writing to the actual index then deleting the row from the indexing queue after it is processed.
Hope it helps,
Joe