Minor issue with MogileFS/Worker/Monitor.pm

dormando dormando at rydia.net
Wed Nov 7 06:42:40 UTC 2007

Fwiw, my only complaint on this tiny patch (aside from being incredibly 
nitpicky) is this:

-    my $num = int(rand 10000);  # this was "$$-$now" before, but we 
don't yet have a cleaner in mogstored for these files
+    my $num = "$$-$now";

I'd prefer to leave the line alone in case of delete suddenly failing on 
someone and prevent the disk from filling up. Unless, of course, there 
are complaints.

Also adds a teeny bit of load, but it's oh so teeny...

(yes, I'm reviewing patches again. Sorry for the huge stack that's built 
up, we'll get through them as soon as we can).


Derek Boonstra wrote:
> Hi,
> The Worker::Monitor probes the devices by putting a test file on the 
> storage.  This is followed by retrieving the data and verifying its 
> content.  The test file on the storage device is then abandoned.
> Though limited to 10000 files, with many trackers running against the 
> storage nodes, the number of files on devX/test-write builds up quickly.  
> The proposed patch attached sends a delete request to the storage node 
> after the put/get test is found to be successful.  In this manner, 
> Monitor cleans up after itself and the task cleaning up is not laid upon 
> mogstored as I have seen with other proposals on this list.

More information about the mogilefs mailing list