So I tried to reproduce your problem.
USed server with similar config: Mgnl 4.3.1, bundle PM, derby, JVM
flags:
-XX:PerfDataMemorySize=256M
-XX:PerfMaxStringConstLength=4096
-XX:ReservedCodeCacheSize=512M
-XX:MaxPermSize=384m
-Xms128M
-Xmx1024M
-Djava.awt.headless=true
-Xdebug
-Xrunjdwp:transport=dt_socket,address=8000,server=y,suspend=n
(you can safely ignore the first half of those as they are just a left
over from some other tests)
Created (hopefully) similar structure with following groovy script:
hm = ctx.getHierarchyManager("website");
test = hm.createContent("/", "test", "mgnl:content");
hm.save();
for (i = 0; i < 20; i++) {
subnode = test.createContent("subnode-" + i, "mgnl:content");
test.save();
for (j = 0; j < 1000; j++) {
subChild = subnode.createContent("subChild-" + i + "-" + j,
"mgnl:contentNode");
for (k = 0; k < 10; k++) {
subChild.setNodeData("test", "the west");
};
subnode.save();
};
println "created " + i + "K nodes";
};
hm.save();
and then tried to delete them via AdminCentral.
Did the same for 10, 20 and 30 thousands nodes.
The result is ... unfortunatelly (or not, depending on the point of
view), I could not reproduce your issue. Furthermore, when observing the
memory, there was no increasing trend, it held steady between 700-800MB
for the duration of deletion and only the time was getting longer with
bigger amount of nodes. CPU spiked few times to 105-110%, but most of
the time held between 30-70%.
So either your structure is more complicated or there are some (big)
binary nodes in your structure or it is something else all together.
Do you have any observation defined on your server that might trigger
other ops while the deletion is in progress? Are there binaries in your
node structure? Can you make a memory dump on OOME and see what kind of
objects occupy the memory?
You might also want to try with some other DB, to rule out derby related
issues for sure as Philipp suggested in his other mail.
Cheers,
Jan
On Mon, 2010-03-29 at 10:03 +0200, Will Scheidegger wrote:
> On 29.03.2010, at 09:49, Jan Haderka wrote:
>
> >
> > On Thu, 2010-03-25 at 15:36 +0100, Will Scheidegger wrote:
> >> Sure, here we go, but I'm afraid it's not very helpful:
> >
> > seems you are right. few more questions tho:
> > - those few thousands nodes are flat or have any sort of hierarchy?
>
> The nodes are in a "mild" hiearchy: 10 main folders with 10 subfolders each.
> The nodes itself have some subnodes too.
>
> > - what kind of PM and what DB are you using?
>
> Standard out-of-the-box (Derby that is, right?).
>
> I tried deleting them programmatically now, but failed miserably too. When I
> looked at the "delete()" method in the AdminTreeMVCHandler class I also
> noticed why: My code was no different than what is done here...
>
> Thanks again for the help, Jan.
>
> -will
>
> ----------------------------------------------------------------
> For list details see
> http://www.magnolia-cms.com/home/community/mailing-lists.html
> To unsubscribe, E-mail to: <[email protected]>
> ----------------------------------------------------------------
----------------------------------------------------------------
For list details see
http://www.magnolia-cms.com/home/community/mailing-lists.html
To unsubscribe, E-mail to: <[email protected]>
----------------------------------------------------------------
