Re: beagles eating my /var filesystem

> All are in the index.  One group is in the filesystem but not tracked by
> dpkg (not really relevant to this issue, admittedly) so not really a
> beagle problem.  The other group, as you can see below, are not even in
> the filesystem but still in the beagle index.  That does seem like a
> beagle problem.

Gotcha. You mentioned that you started using --enable-deletion recently - so this is what is happening. I will try to be brief.
The way the enable-deletions work is, while traversing the directories, beagle keeps a note of all the directories that were modified since last index. When indexing is done, it starts looking at those directories (it prints a message "Checking xxx for deleted files and directories") and recursively finds files and directories in them which are deleted. The last step involves getting a list of all files in that directory and checking if they exist, and if not deleting them - basically pretty expensive. In short, only the dirty directories are checked for deleted content (directories within deleted directories are handled on the way).

So, if you sometimes run without --enable-deletion, beagle will not remove the deleted content but obviously would mark every directory as clean after the indexing is over. So the next time you run with --enable-deletion, it sees the directory is clean and will not check it for deleted content.

I think there should be a --enable-force-deletion to just scan all the uris in the index and remove the ones not in the index - it will be pretty expensive but users can run it once in a while.

There might still be bugs in enable-deletion, as I said it never got enough testing. One thing you could do for testing is, install some doc package so that /usr/share/doc gets dirty again, run with --enable-deletion and see if /usr/share/doc/bzr and its recursive contents are removed. Look for the "Checking xxx for deleted files and directories" message near the end of the output messages.

Unfortunately, all of the above is probably a red herring wrt the original problem. Those extra entries cannot occupy that amount of space. Its what I suspect a problem in Lucene and the workaround I added in 0.3.0 might fix such problems. I will confirm once I get a chance to look inside the segments file.

- dBera

Debajyoti Bera @
beagle / KDE fan
Mandriva / Inspiron-1100 user

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]