Good. Manually replacing the tmp dir to current works. It's not corrupt or whatsoever and is readable. I recall this was not the case for linkdb but i may have don't something else wrong.
Of course i had to remove the lock file as well. > The same now occasionally happens with the crawldb. Today we tested new > filters a couple of times. The job failed without failure twice, no strange > error message or whatever. The only clue is that the tmp files does not > replace the `current` dir. > > Anything? > > > Hi, > > > > Last night i started another round of link inverting and merging with the > > existing LinkDB. Many hours later i see the job finished without failures > > or errors. > > > > Yet the LinkDB wasn't updated. To my surprise i saw the intermediate > > files hanging around in my Hadoop ~ but i expected the temp files to be > > cleaned and the new LinkDB file to be copied over the old one. > > > > The new file that ought to be copied over is also broken. I cannot > > readlinkdb. Gladly the old still works but before i try to debug some > > process that takes many hours.. any idea? > > > > Thanks
