[freenet-support] Freenet 0.7 build 1178 (and 1176, 1177)

[Dennis Nezic]

On Fri, 14 Nov 2008 18:21:50 -0500, Dennis Nezic
 wrote:

> On Fri, 14 Nov 2008 19:39:45 +, Matthew Toseland
>  wrote:
> 
> > Freenet 0.7 build 1178 is now available.

I spoke too soon. It works fine for an hour or two, but then, at least
for me, all communcation (net input/output) STOPS. (Well, ~100Bytes/sec
both ways--but, for all intents and purposes, it's offline.)

After almost exactly 3 hours, it begins Announcing itself
(successfully, apparently) although nothing really changes.

And this is repeatable. Restarting the node leads to the exact same
behaviour.

There are very few errors in the freenet logs. A couple per hour,
maybe. But, the "common" one that I spotted freenet.node.PeerManager,
UdpSocketHandler ERROR addPeerNodeStatus(): "node already in
peerNodeStatuses" ... "status DISCONNECTED". But, it's ip address is
still listed as CONNECTED, and data is still being transferred with it,
although the port number in the log file is different than the one that
is listed as connected. (I'm not sure if this is related.)

Re: [freenet-support] Freenet 0.7 build 1178 (and 1176, 1177)

[Dennis Nezic]

On Fri, 14 Nov 2008 18:21:50 -0500, Dennis Nezic
[EMAIL PROTECTED] wrote:

 On Fri, 14 Nov 2008 19:39:45 +, Matthew Toseland
 [EMAIL PROTECTED] wrote:
 
  Freenet 0.7 build 1178 is now available.

I spoke too soon. It works fine for an hour or two, but then, at least
for me, all communcation (net input/output) STOPS. (Well, ~100Bytes/sec
both ways--but, for all intents and purposes, it's offline.)

After almost exactly 3 hours, it begins Announcing itself
(successfully, apparently) although nothing really changes.

And this is repeatable. Restarting the node leads to the exact same
behaviour.

There are very few errors in the freenet logs. A couple per hour,
maybe. But, the common one that I spotted freenet.node.PeerManager,
UdpSocketHandler ERROR addPeerNodeStatus(): node already in
peerNodeStatuses ... status DISCONNECTED. But, it's ip address is
still listed as CONNECTED, and data is still being transferred with it,
although the port number in the log file is different than the one that
is listed as connected. (I'm not sure if this is related.)
___
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:[EMAIL PROTECTED]

[freenet-support] Freenet 0.7 build 1178 (and 1176, 1177)

[Matthew Toseland]

On Friday 14 November 2008 23:21, Dennis Nezic wrote:
> On Fri, 14 Nov 2008 19:39:45 +, Matthew Toseland
>  wrote:
> 
> > Freenet 0.7 build 1178 is now available.
> 
> Wow! What a difference. 1178 is like a totally different program, with
> respect to both cpu and memory usage. THANK YOU! It's no longer peaking
> for minutes at a time. It's consistently low. My 15minute load average
> is below 1 now! Hurray. Memory use also seems much lower.
> 
> Regarding the "not acked packets" problem, I have alrady been alerted
> about 5 peers with "severe" problems, one of which is a darknet peer. 2
> of the others are listed as CONNECTED. And the other 2 are listed as
> BACKED OFF. All of them are running 1178.
> 
> Also, I should repeat ;), One of my peers (ip addresses) is CONNECTED
> to me both in opennet and darknet.

That's to be expected. We kill the connection, and purge the trackers etc, but 
we can reconnect afterwards. Also, it's quite possible that those peers were 
running an earlier version when they actually broke. However, it's also 
possible that we still have the underlying bug... I've not made much headway 
with finding it, next week we should know whether it's still active, since 
1178 is mandatory on Sunday...
> 
> All in all, though, 1178 is a HUGE improvement!
-- next part --
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: not available
URL: 

[freenet-support] Freenet 0.7 build 1178 (and 1176, 1177)

[Matthew Toseland]

Freenet 0.7 build 1178 is now available. Please upgrade. Sorry I accidentally 
released 1176, then had some bugfixes in 1177 and 1178. Anyway, the main 
changes relative to 1175:
- Try to prevent both sides of a connection attempting to rekey at the same 
time, by having the receiver start 2.5 minutes early. This seems to have been 
the root cause of many of the recent problems. It may not be completely 
fixed, but should happen much less often.
- When we disconnect from a peer because of it not acknowledging our packets, 
tell the user via a useralert as well as logging an error.
- Get rid of WouldBlockException errors as well: If we cannot send a packet to 
a peer because there are too many packets in flight and none are being 
acknowledged, don't busy-loop and spam the logs, wait for it to become 
un-blocked again. If it doesn't become un-blocked in 10 minutes, log an error 
and tell the user via a useralert.
- Change the default log rotation period to 10 minutes rather than 1 hour to 
save disk space.
- Clarify the config option for packet padding: Turning off packet padding is 
BAD. Packet padding is essential against a passive attacker.

In the next few days you may see the "Probably a bug" errors on the web 
interface. Hopefully once this build is mandatory (on Monday) these will go 
away. If you still see these from Tuesday onwards, let us know.

Please update, and let us know how it works out. Hopefully this build will fix 
most of the recent network problems, oversized log files and high CPU usage 
when not doing anything. If you find any bugs, please report them on the bug 
tracker: https://bugs.freenetproject.org/ . Thanks.
-- next part --
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: not available
URL: 

[freenet-support] Freenet 0.7 build 1178 (and 1176, 1177)

[Dennis Nezic]

On Fri, 14 Nov 2008 19:39:45 +, Matthew Toseland
 wrote:

> Freenet 0.7 build 1178 is now available.

Wow! What a difference. 1178 is like a totally different program, with
respect to both cpu and memory usage. THANK YOU! It's no longer peaking
for minutes at a time. It's consistently low. My 15minute load average
is below 1 now! Hurray. Memory use also seems much lower.

Regarding the "not acked packets" problem, I have alrady been alerted
about 5 peers with "severe" problems, one of which is a darknet peer. 2
of the others are listed as CONNECTED. And the other 2 are listed as
BACKED OFF. All of them are running 1178.

Also, I should repeat ;), One of my peers (ip addresses) is CONNECTED
to me both in opennet and darknet.

All in all, though, 1178 is a HUGE improvement!

[freenet-support] Freenet 0.7 build 1178 (and 1176, 1177)

[Matthew Toseland]

Freenet 0.7 build 1178 is now available. Please upgrade. Sorry I accidentally 
released 1176, then had some bugfixes in 1177 and 1178. Anyway, the main 
changes relative to 1175:
- Try to prevent both sides of a connection attempting to rekey at the same 
time, by having the receiver start 2.5 minutes early. This seems to have been 
the root cause of many of the recent problems. It may not be completely 
fixed, but should happen much less often.
- When we disconnect from a peer because of it not acknowledging our packets, 
tell the user via a useralert as well as logging an error.
- Get rid of WouldBlockException errors as well: If we cannot send a packet to 
a peer because there are too many packets in flight and none are being 
acknowledged, don't busy-loop and spam the logs, wait for it to become 
un-blocked again. If it doesn't become un-blocked in 10 minutes, log an error 
and tell the user via a useralert.
- Change the default log rotation period to 10 minutes rather than 1 hour to 
save disk space.
- Clarify the config option for packet padding: Turning off packet padding is 
BAD. Packet padding is essential against a passive attacker.

In the next few days you may see the Probably a bug errors on the web 
interface. Hopefully once this build is mandatory (on Monday) these will go 
away. If you still see these from Tuesday onwards, let us know.

Please update, and let us know how it works out. Hopefully this build will fix 
most of the recent network problems, oversized log files and high CPU usage 
when not doing anything. If you find any bugs, please report them on the bug 
tracker: https://bugs.freenetproject.org/ . Thanks.


pgpREOBv5UTIk.pgp
Description: PGP signature
___
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:[EMAIL PROTECTED]

Re: [freenet-support] Freenet 0.7 build 1178 (and 1176, 1177)

[Dennis Nezic]

On Fri, 14 Nov 2008 19:39:45 +, Matthew Toseland
[EMAIL PROTECTED] wrote:

 Freenet 0.7 build 1178 is now available.

Wow! What a difference. 1178 is like a totally different program, with
respect to both cpu and memory usage. THANK YOU! It's no longer peaking
for minutes at a time. It's consistently low. My 15minute load average
is below 1 now! Hurray. Memory use also seems much lower.

Regarding the not acked packets problem, I have alrady been alerted
about 5 peers with severe problems, one of which is a darknet peer. 2
of the others are listed as CONNECTED. And the other 2 are listed as
BACKED OFF. All of them are running 1178.

Also, I should repeat ;), One of my peers (ip addresses) is CONNECTED
to me both in opennet and darknet.

All in all, though, 1178 is a HUGE improvement!
___
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:[EMAIL PROTECTED]

Re: [freenet-support] Freenet 0.7 build 1178 (and 1176, 1177)

[Matthew Toseland]

On Friday 14 November 2008 23:21, Dennis Nezic wrote:
 On Fri, 14 Nov 2008 19:39:45 +, Matthew Toseland
 [EMAIL PROTECTED] wrote:
 
  Freenet 0.7 build 1178 is now available.
 
 Wow! What a difference. 1178 is like a totally different program, with
 respect to both cpu and memory usage. THANK YOU! It's no longer peaking
 for minutes at a time. It's consistently low. My 15minute load average
 is below 1 now! Hurray. Memory use also seems much lower.
 
 Regarding the not acked packets problem, I have alrady been alerted
 about 5 peers with severe problems, one of which is a darknet peer. 2
 of the others are listed as CONNECTED. And the other 2 are listed as
 BACKED OFF. All of them are running 1178.
 
 Also, I should repeat ;), One of my peers (ip addresses) is CONNECTED
 to me both in opennet and darknet.

That's to be expected. We kill the connection, and purge the trackers etc, but 
we can reconnect afterwards. Also, it's quite possible that those peers were 
running an earlier version when they actually broke. However, it's also 
possible that we still have the underlying bug... I've not made much headway 
with finding it, next week we should know whether it's still active, since 
1178 is mandatory on Sunday...
 
 All in all, though, 1178 is a HUGE improvement!


pgpRpMyPP1lGK.pgp
Description: PGP signature
___
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:[EMAIL PROTECTED]