On Thursday 11 November 2010 19:36:02 l...@hush.com wrote: > > On Wed, 10 Nov 2010 14:16:01 -0600 Matthew Toseland > <t...@amphibian.dyndns.org> wrote: > >On Wednesday 10 November 2010 18:42:48 l...@hush.com wrote: > >> > >> On Tue, 09 Nov 2010 19:13:12 -0600 Matthew Toseland > >> <t...@amphibian.dyndns.org> wrote: > >> >On Tuesday 02 November 2010 17:23:46 l...@hush.com wrote: > >> >> Tried inserting a small directory through Fproxy, seems to > >> >encode > >> >> the blocks and then stops with 0 blocks inserted and sits. I > >> >think > >> >> this is the error message. Tries several times as SSK and > >USK. > >> > > >> >How exactly did you upload it? Did you choose the third option > >and > >> >specify a particular key? > >> > >> Yes, to a specific key. It did eventually work, but took much > >> longer than through jsite or TMCI. It was small about 10 Meg. > >Just > >> hadn't seen anything stall at 0 blocks overnight. > >> > >> I have sucessfuly inserted smallish freesites through frproxy > >but > >> every attempt at a large one 1-4Gig mirror of sites of a type > >you > >> have mentioned have always stalled at some point. > > > >Unfortunately our support for big sites is quite limited and > >inefficient at the moment. A site with some big files won't be a > >problem but a site with a lot of smallish files will be inserted > >very inefficiently and thus take ages and fall out too quickly. We > >hope to resolve this by enabling saces' code (you can do that > >yourself if you feel like it), but it is not yet fully debugged > >and doesn't support persistent uploads at present. > > Thats it, a couple of the sites have 200,000-500,000 small files. > > One did work with jsite about 3G and 225,000 files and took about 5 > days with other things going on as well, but the site itself was > poorly designed and wouldn't display properly, fault of the site > designer. > >> > >> Have tried that thru jsite and tmci but it takes about a week > >and > >> an uninterrupted week is a very rare thing. > >> > >> I'll give it another shot when things settle down a bit. > > > >It should not take a week to upload a 10MB freesite! My blog is > >35MB and it takes about half an hour. Granted most of the size is > >biggish files, the HTML is quite small. > > > It was a week for the 3 Gig. > Though even 600 blocks site insert thru fproxy takes much longer > than either jsite or TCMI.
Possibly priority problems, which means other inserts running? Or just other requests, chat clients etc, if it stalls at the end? > > >And yes, stalling at 0 blocks overnight is bad. > > > >Was there anything else running? Uploads of individual files, > >maybe at high priority? Client apps? Etc? > > No client apps, my simplest node no even running most plugins, a > few inserts at med and high. That rules out the stalling at the end theory. However it could still be a priority problem if the freesite was at a lower priority, or if it started there and there is a problem changing priorities? > > >Is your node generally behaving itself - can you access freesites > >reasonably quickly? Does your node show any obvious errors e.g. > >high ping time? What is your nodeAveragePingTime on the stats > >page? The top reasons for preemptive request rejection (local and > >not)? The pInstantReject? The bandwidth usage? > > That's all fine. My problem atm is growing persistent blob. I have > two nodes and I have to clear what I want inserted and downloaded > before I can delete the ever growing blobs and clean that up. Do you get more blob leaks due to inserts or requests? Can you find out? > On > one upload q, I can't even delete anything from it but I can change > priority to paused so a nod's as good as a wink. The other kept > getting wrapper timeouts while encoding upload and doing store > cleaning. Slow computer? Swapping, or maybe low memory limit set for Freenet? > When it came up again just said "internal" error when > trying to display either queue through Fproxy. Defragging twice > somehow brought both queues back. Did it run out of disk space, suffer a system reset or similar problem that would cause data corruption? > > I notice changing the priority back and forth on things that look > stalled or slow seems to give them a kick to wake up. Only on requests? Or on inserts too? It doesn't matter what priority you set them to? You should grep the logs for obvious ERROR's. They might tell me something interesting. Try something like: zgrep "\(NullPointer\)\|\(Illegal\)\|\(ArrayIndex\)\|\(client.*ERROR\)\|\(Internal [Ee]rror\)" logs/freenet-* > > All and all it ain't all bad. I'm just going to wait until the q's > are clear on one node to attempt a large persistent site insert, > also depending on where your work is at the time.
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ Support mailing list Support@freenetproject.org http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:support-requ...@freenetproject.org?subject=unsubscribe
