63880 - News hangs downloading (expired?) articles

Status: VERIFIED WORKSFORME

(MailNews Core :: Networking: NNTP, defect)

Description

[Peter Janes]

Fairly often (but not reliably reproducible), news hangs (throbber and status
bar animations continue) while downloading messages on news1.sympatico.ca.  It
happens most often when an article that's been expired is requested, but also
happens on other valid articles.  A partial NNTP log is included here:

1024[80587e8]: Sending: GROUP alt.fan.harlan-ellison
1024[80587e8]: Next state: NNTP_RESPONSE
1024[80587e8]: Receiving: 211 25 24506 24540 alt.fan.harlan-ellison
1024[80587e8]: Next state: NNTP_SEND_GROUP_FOR_ARTICLE_RESPONSE
1024[80587e8]: Next state: NNTP_SEND_ARTICLE_NUMBER
1024[80587e8]: Sending: ARTICLE 24526

After 10 minutes or so, I got:

1024[80587e8]: Next state: NEWS_FREE

Neither the message or an error is displayed during this time, and the message
pane is "frozen" (i.e. it does not paint itself, just keeps the contents of any
other windows that have been over it).  I can navigate in the folder and thread
panes, though.

Navigating to another (expired) message, I get:

1024[80587e8]: Next state: SEND_FIRST_NNTP_COMMAND
1024[80587e8]: Next state: NNTP_SEND_GROUP_FOR_ARTICLE
1024[80587e8]: Sending: GROUP alt.fan.harlan-ellison
1024[80587e8]: Next state: NNTP_RESPONSE
1024[80587e8]: Receiving: 211 11 24528 24540 alt.fan.harlan-ellison
1024[80587e8]: Next state: NNTP_SEND_GROUP_FOR_ARTICLE_RESPONSE
1024[80587e8]: Next state: NNTP_SEND_ARTICLE_NUMBER
1024[80587e8]: Sending: ARTICLE 24523
1024[80587e8]: Next state: NNTP_RESPONSE
1024[80587e8]: Receiving: 423 No Such Article In Group
1024[80587e8]: Next state: SEND_FIRST_NNTP_COMMAND_RESPONSE
1024[80587e8]: Next state: NNTP_ERROR
1024[80587e8]: Next state: NEWS_FREE

Sometimes I can navigate away from an unexpired message, then select it again
and it will be retrieved.  Other times it will remain unavailable for the
session or across sessions.

I always get a 423 return code if I try to view the expired articles myself, as
with the second (successful) log excerpt above:

> telnet news1.sympatico.ca 119
Trying 206.172.150.50...
Connected to news1.sympatico.ca.
Escape character is '^]'.
200 Bell Nexxia News Server -- news://news1.sympatico.ca/ (Typhoon v1.2.2)
GROUP alt.fan.harlan-ellison
211 11 24528 24540 alt.fan.harlan-ellison
ARTICLE 24526
423 No Such Article In Group
QUIT
205 GoodBye
Connection closed by foreign host.

Comment 1

[Peter Janes]

Linux build 2000122710.  Also, I don't have this problem at all reading
news.mozilla.org.

Comment 2

[[e-mail] (gone - use [e-mail] instead)]

Is this an especially slow server?  

Comment 3

[Peter Janes]

It doesn't seem to be--how would I go about measuring the speed?

The Ellison group doesn't have any particularly large posts, so it shouldn't be
timing out while retrieving them.

Comment 4

[[e-mail] (gone - use [e-mail] instead)]

There doesn't seem to be anything out of the ordinary with your setup, so I can
only assume there is some kind of underlying network issue here.  I'll try it
with that newsgroup on my P133 and see if I get any results.

Comment 5

[[e-mail] (gone - use [e-mail] instead)]

Hrm, strange, this works fine for me (there was only 26 postings when I loaded
this ng).  I used RedHat 7, build 2001011208.

Comment 6

[Peter Janes]

Sounds like it may be the Sympatico server, then.  Any suggestions on debugging?

Comment 7

[[e-mail] (gone - use [e-mail] instead)]

Not sure how much David or Seth know about Typhoon servers.  Adding David to the
list.  Are you on dialup?  What is the last build you tried this with? 

Comment 8

[Peter Janes]

I'm running 2001013108 on "high-speed" DSL.  I haven't seen the problem as much
recently (since the "only one download thread for news" bug was fixed), but it
still occurs occasionally.  There hasn't been as much traffic in the group
recently though, and I'm not really subscribed to much else on that server--I'll
try a couple of higher-traffic ng's and see if I see it more.

Comment 9

[[e-mail] (gone - use [e-mail] instead)]

Okay, if you find an especially high traffic newsgroup, please post it here. 
They are very useful for stress testing!

Comment 10

[Peter Janes]

Trying with a 500-message set of rec.arts.sf.tv.babylon5.moderated in build
2001020908, I've gotten a few "freezes", but they have gone away after selecting
one or two messages, then returning to the "frozen" message.  I've read about
250 messages and I'd say I've frozen on 15 of those; three occurred in a row,
and the rest have been single occurrences.  The freezes occurred, about 95% of
the time, after I went back to a post I had already read (to see some thread
context that wasn't quoted), then continued to the 'next unread'.

Regarding the comment on the server being slow, I got about 1.5K/s downloading
the list of newsgroups (on a 128K DSL connection).  So I guess that's slow. :-)
 http://www.sympaticousers.org/ has some discussion of the poor news servers,
but nothing specifically about performance.

Comment 11

[[e-mail] (gone - use [e-mail] instead)]

Wow, hey Seth, take a look at this,
http://www.sympaticousers.org/action/news_service.htm
Looks like they have lots of issues with headers not being downloaded, etc.

Comment 12

[Peter Janes]

Hmm, I must be reading that page differently.  I thought it referred mostly to
headers (and messages) not being synchronized between the different servers...
although they do mention Netscape (I assume 4.x) having problems with
out-of-order headers.  I haven't seen any problems doing a header download when
entering the newsgroups.

Comment 13

[[e-mail] (gone - use [e-mail] instead)]

Oh, I haven't read the whole article (yet).  I'm referring to the screenshots, 
where it shows "Warnings" next to Downloading news articles and "Failed" 
checking (what I assume to be) "Bell" a server for new messages.  Notice, there 
don't seem to be any updated warning for this year, though it seems they've had 
their share of problems.

Comment 14

[[e-mail] (gone - use [e-mail] instead)]

Here's a link to the type of server that your using:
http://www.bcandid.com/discussion/service_providers/typhoon/typhoon_content.html

Comment 15

[Jens Thiele]

mozilla hangs sometimes/often when using leafnode as newsserver

i think this bug is related to 63880 (or perhaps 58247)

unfortunately it works if I connect directly to the news server (not using leafnode)
and mozilla doesn't always hang if using leafnode. It only hangs on some groups
but if it hangs it will always do.

tried moving the msf file as described in bug 58247
but it still happened (the file was empty btw)

unfortunately i don't know anything about the nntp protocol
but i think mozilla is requesting the article with id 1
that doesn't exist on the server (doesn't exist in the leafnode spool directory)
and perhaps this is the problem ?

so i would say this is bug 63880 ?

i coppied a news message (2) to 1 => now the requested message 1 exists => it works
(unfortunately it seems mozilla doesn't know how many messages there are now but
this is not related to this bug i think.)

the problem doesn't occur with mozilla 0.8.0
or other nntp clients (pine/netscape, navigator 4.7?)

another idea: perhaps it is a problem with "remove all expired articles"


If mozilla hangs there is no network traffic and send and receive buffers are empty.
Leafnode is waiting for incoming traffic.

if you want i can add a lot of detail
(gdb stacktrace , traffic sniffed, netstat, ...)

My system details:My system:
debian gnu/linux 2.2 potato (infofficial x 4.0.2 packages for potato)
linux 2.4.2
mozilla 0.9
leafnode 1.9.9

jens@quark:/etc > ldd /usr/local/mozilla/mozilla-bin
	libgkgfx.so => not found
	libxpcom.so => not found
	libmozjs.so => not found
	libjsj.so => not found
	libplds4.so => /usr/lib/libplds4.so (0x40020000)
	libplc4.so => /usr/lib/libplc4.so (0x40024000)
	libnspr4.so => /usr/lib/libnspr4.so (0x40029000)
	libpthread.so.0 => /lib/libpthread.so.0 (0x4006b000)
	libdl.so.2 => /lib/libdl.so.2 (0x4007e000)
	libc.so.6 => /lib/libc.so.6 (0x40083000)
	libstdc++-libc6.1-1.so.2 => /usr/lib/libstdc++-libc6.1-1.so.2 (0x40160000)
	libm.so.6 => /lib/libm.so.6 (0x401a2000)
	/lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)

lib/ld-linux.so.2 -> ld-2.1.3.so
lib/libc.so.6 -> libc-2.1.3.so
lib/libdl.so.2 -> libdl-2.1.3.so
lib/libm.so.6 -> libm-2.1.3.so
lib/libpthread.so.0 -> libpthread-0.8.so
usr/lib/libstdc++-libc6.1-1.so.2 -> libstdc++-2-libc6.1-1-2.9.0.so

Comment 16

[Jens Thiele]

Attachment: debugging information (stacktrace, netstat, traffic sniffed)

Comment 17

[Jens Thiele]

I reported the problem with leafnode. Since mozilla 0.9.3 it seems to be fixed?

Comment 18

[[e-mail] (gone - use [e-mail] instead)]

Peter Janes, is this bug gone for you too, now?

Comment 19

[Peter Janes]

I've only noticed "hangs" once or twice in the last month, and they're
attributable to outside influences, so I'd say it's fixed.

Comment 20

[[e-mail] (gone - use [e-mail] instead)]

Okay, marking as such.  One of you guys please verify then (I'm not running a
leafnode server, currently).  Thanks.

Comment 21

[[e-mail] (gone - use [e-mail] instead)]

verified.  Two people's word is good enough for me (especially since Peter was 
the reporter).