[leafnode-list] Excessive use of memory - no2

[giuliox@xxxxxx (Giulio Orsero)]

Leafnode 1.9.2, linux 2.0.33.

The title says no2 because I've already posted a while ago about a problem
with fetch going out of control and swapping for about 60MB and more.

Conrelius said:

	This could be due to gaps in the numbering of articles. When you have
	two articles in a group numbered 5 and 138429, leafnode will try to
	allocate an array of (138429-5+1) elements. (The same problem is also
	present in texpire.)

	There is no obvious solution to this. One thing you could try is to
	renumber the articles which definitely doesn't sound like fun.

An it was true. There was a group with weird numbering; I resetted it by hand
and all was well again.

Just to  be sure I added some "ulimit's" in the script that launches fetch,
hoping not to see again something like that.

This time happened with the leafnode process.....
The swap went 80MB, and I had to kill it.
Now I'm going to ulimit leafnode too.

I know you are busy, but wouldn't it be possible for you, Cornelius, to make a
simple script that checks the spool for abnormal numbering (I'd do it myself
If I knew how leafnode works, but unfortunately I don't know anything about c)
and reports that there's a problem, so that one can routinely  launch it
before running leafnode/fetch/texpire and prevent these problems?

Or could you suggest some secure ulimit settings (on linux I can't use ulimit
-v, for now I use a combination of -d -m -s) that can let leafnode do its work
but block it when it has problems?


Thanks.

-- 
Giulio
giuliox@xxxxxx

-- 
leafnode-list@xxxxxxxxxxxxxxxxxxxxxxxxxxxx -- mailing list for leafnode
To unsubscribe, send mail with "unsubscribe" in the subject to the list