Re: 3TiB-Node: Installation fails with Java out of memory error

Dennis New <[email protected]>
Newsgroups gmane.network.freenet.support
Message-ID <[email protected]>
On Thu, 17 Apr 2014 10:52:40 +0200, Wolfram Goetz wrote:
> > Gesendet: Mittwoch, 16. April 2014 um 04:10 Uhr
> > Von: "Dennis New" <[email protected]>
> > An: [email protected]
> > Cc: "Wolfram Goetz" <[email protected]>
> > Betreff: Re: [freenet-support] 3TiB-Node: Installation fails with
> > Java out of memory error
> >
> > On Mon, 14 Apr 2014 21:05:31 +0200, Wolfram Goetz wrote:
> > > Hello everyone,
> > > 
> > > I had successfully run a 400GiB-Node in a Windows 7 x64
> > > environment, running as a virtual machine in VirtualBox with 6GB
> > > RAM. Now I moved the whole installation to a larger harddisk and
> > > tried to increase the datastore size to 3TiB. This results in a
> > > java out of memory error. This is from my wrapper.log:
> > > 
> > > INFO   | jvm 1    | 2014/04/14 17:08:17 | Initializing SSK
> > > Datastore (46033289 keys)
> > > INFO   | jvm 1    | 2014/04/14 17:08:17 | WrapperManager Error:
> > > Error in WrapperListener.start callback.
> > > java.lang.OutOfMemoryError: Java heap space STATUS | wrapper  |
> > > 2014/04/14 17:08:17 | Filter trigger matched. Restarting JVM.
> > > INFO   | jvm 1    | 2014/04/14 17:08:17 | WrapperManager Error:
> > > java.lang.OutOfMemoryError: Java heap space
> > > STATUS | wrapper  | 2014/04/14 17:08:17 | Filter trigger matched.
> > > Restarting JVM.
> > > 
> > > After a few more error messages freenet shuts down completely.
> > > Same problem with a new & clean install of freenet. 
> > > 
> > > Memory settings from my wrapper conf:
> > > 
> > > wrapper.java.initmemory=256
> > > wrapper.java.additional.4=-XX:MaxPermSize=256m
> > > wrapper.java.maxmemory=1024
> > > 
> > > Did not work with the original settings, too. Any idea what I
> > > could do?
> > 
> > Perhaps try increasing your MaxPermSize, to something like 500M or
> > so? And check (in fproxy's Statistics page) that you aren't
> > reaching your 1gig maxmemory, which should probably be enough?
> 
> Thanks for the answer. Played around with the settings on a fresh
> freenet install, but I do not really know what I'm doing... This
> seems to work more or less:
> 
> wrapper.java.initmemory=60 (default)
> wrapper.java.additional.4=256

I'm pretty sure that's invalid and being ignored. It's supposed to be
additional command line arguments that you would pass to "java"... i.e.
they should start with a dash (-).

> wrapper.java.maxmemory=1536

That's probably what fixed things. Before you were using 1gig. Now
it's using 1.3gig.


> On the stats page, 1,45 GiB are allocated for Java and 1,30 GiB are
> in use. I left alone the MaxPermSize setting, which seems to point to
> another variable:
> 
> MaxPermSize=wrapper.app.parameter.1=freenet.ini

This also seems to be an invalid line. I think all the lines in the
wrapper.conf should start with "wrapper.*". Moreoever, freenet.ini does
not / should not mention these lower level jvm settings (like
MaxPermSize). It might be worthwhile to paste your wrapper.conf file to
see if there's other bogus stuff.


> Still testing, had some additional stability problems because
> VirtualBox seems to have problems (BSODs with hardware virtualization
> and two CPUs in den VM). I successfully started a copy of the
> original 400GiB node and set the datastore size to 3000GiB. But after
> half an hour it still says "Datastore(CHK-cache) resize in progress:
> 0/6137772". Hmmm. I'll check back later.

That sounds normal for datastore resizing.
_______________________________________________
Support mailing list
[email protected]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:support-request-RdDMkVZAZeuJnvDnx1genB2eb7JE58TQ@public.gmane.org?subject=unsubscribe
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.