I'll try to refrain from chastising you for using a log as a work queue... :-)
but I suppose the "cleanest" way would be to log to a database and then delete those rows once the files have been indexed. Alternatively, if that adds too many moving parts and you want to stick with text files, leave a space in the log records and rewrite them with a marker (e.g. "*") in that space once the files have be indexed to mark them as "deleted" or "handled" as the case may be. On Thu, 2011-12-15 at 08:59 +0100, Michael Wechner wrote: > Hi > > As some of you might now Apache Lucene creates a "write.lock" when an > IndexWriter is updating the index, > hence it can happen that another IndexWriter (inside another thread) > won't be able to update the same index at the same time and based on a > specific timeout something might not get indexed. > > We are currently logging all such instances when something does not get > indexed. > > Now based on that log file, containing hints re files which have not > been indexed yet, we would like to introduce > a process starting to re-index these specific files and if successful, > then we would like to remove the corresponding particular entry from the > log file. > > I understand that log4j might not be intended for such a usecase and > maybe we shouldn't try to misuse it, but I would appreciate any pointers > or ideas or alternatives. > > Thanks > > Michael > > --------------------------------------------------------------------- > To unsubscribe, e-mail: log4j-user-unsubscr...@logging.apache.org > For additional commands, e-mail: log4j-user-h...@logging.apache.org >
signature.asc
Description: This is a digitally signed message part
