Re: Anecdotal: High IO when unable to connect to peers

On Saturday 06 Aug 2011 17:36:29 Matthew Toseland wrote:
> On Wednesday 27 Jul 2011 14:20:59 Dennis Nezic wrote:
> > This is somewhat anecdotal and sparse on details, but probably something
> > significant. I have noticed that whenever I have trouble connecting to
> > peers, ie. when my version becomes "too old" to connect to the
> > mainstream nodes (I update my node manually -- gentoo package), the disk
> > IO skyrockets and my computer becomes bogged down (uptime values of 15
> > +), and I'm forced to kill the node. With the latest mandatory 1388,
> > (and my 1386), I'm actually not exactly sure if it's a peer-connectivity
> > (with some kind of disk-io hook? -- perhaps related to the node trying
> > to fetch the newer jars), or if it's normal while the node is starting
> > up. Either way, I can't start 1386 now, due to really high IO. I'll
> > test again when I get 1388 to install.
> 
> It is quite possible that it would do more disk I/O when there are such problems (it writes the peers file way too much), but I have no idea how that would cause a system load of 15 (as those writes should be serialised so only one at a time).
> 
> A thread dump when it is doing this would be interesting.
> 
Current master code should write the peers file a lot less, it would be interesting to re-test. (I will release a snapshot at some point or you can compile it).

_______________________________________________
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

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)

iEYEABEIAAYFAk5O1+4ACgkQYUNbc3WUHYi4RQCfRD/AZ4Ns39ZOcifJv9m5P8lD
vUoAoKLEVyPYOWuuATy5A/rdBCJ8LTJW
=t1P/
-----END PGP SIGNATURE-----