Not to hijack your thread, but I'm actually having similar problems with UNS - not even on the SSL server!
It constantly hangs on random articles (selected for d/l from headers that I retrieved from UNS moments ago - ie. not from the global search).
The worst part is that it inevitably leads to streaming errors about having too many connections. I have connections set to 5 (they allow 10) and my firewall always shows fewer than 10 when I recieve this error. UE also only shows the right number of connections in the task manager.
I think UNS is for crap now, and will be switching, but for the time being are there some recommended settings so I can leave downloads unnatended for more than 20 minutes?
I do have some quesion to UE as well... the timeout does not seem to be going well. I have it set to 60 sec read, and I've watched the same 5 stuck tasks say 'connecting' now for 10 minutes. Is that the right setting?
Thanks
I'm currently trying disabling 'keep connections alive' and have a 60 second read timeout. v1.9.9.7
UNS hanging non-SSL connections?
i put your post into a separate topic, since the thread was only relevant for server ungracefully dropping SSL connections with the regard to OpenSSL library which had been subsequently taken care of.
do you have "resolve dns on every connect" in properties->general unchecked? if it is checked - uncheck it and try again.
if you still have connection hanged for more than timeout interval, it might be the connect call, but i'm not sure it is UNS specific, it might be your firewall as well.
if it is not dns resolution, does it give connect error then after some time? how much time? (connect error is not retriable but if you added it to retriable errors yourself set retries to zero).
as to connect i initially added connect timeout code when we explored the SSL timeout error but then i commented it out (still in the source code, but not included), since noone mentioned problems with connect hanging indefinitely, i also checked google groups with no definite conclusion it might be an issue, so i decided it is not a needed option and we better trust the system winsock layer to time out the connect call (when a winsock call is not needed better not to include it since sometimes non-typical solutions may come into "conflict" with firewalls triggering their implementation bugs, you never know).
http://www.netwu.com/newspro/phpBB2/vie ... php?t=2474
if the connect times out reasonably by itself it shouldn't be a problem.
with UNS i'm not quite sure what is happening, it appears they redesigned their website lately making it smaller, duller and dropping their recommended newsreader list alltogether. i've heard usenet.com was hit with record industry law suit in October, i don't know if it is related. if UNS quality got worse as well it shouldn't be related then, maybe shrinking user base and they had to fire staff/downgrade, who knows, e.g. at one point e.g. newsfeeds.com apparently had been restructured so it is not unthinkable. i got reports by several UNS users that "no such article" bug connected to faulty message-id index resurfaced again, don't know if they fixed it.
do you have "resolve dns on every connect" in properties->general unchecked? if it is checked - uncheck it and try again.
if you still have connection hanged for more than timeout interval, it might be the connect call, but i'm not sure it is UNS specific, it might be your firewall as well.
if it is not dns resolution, does it give connect error then after some time? how much time? (connect error is not retriable but if you added it to retriable errors yourself set retries to zero).
as to connect i initially added connect timeout code when we explored the SSL timeout error but then i commented it out (still in the source code, but not included), since noone mentioned problems with connect hanging indefinitely, i also checked google groups with no definite conclusion it might be an issue, so i decided it is not a needed option and we better trust the system winsock layer to time out the connect call (when a winsock call is not needed better not to include it since sometimes non-typical solutions may come into "conflict" with firewalls triggering their implementation bugs, you never know).
http://www.netwu.com/newspro/phpBB2/vie ... php?t=2474
if the connect times out reasonably by itself it shouldn't be a problem.
with UNS i'm not quite sure what is happening, it appears they redesigned their website lately making it smaller, duller and dropping their recommended newsreader list alltogether. i've heard usenet.com was hit with record industry law suit in October, i don't know if it is related. if UNS quality got worse as well it shouldn't be related then, maybe shrinking user base and they had to fire staff/downgrade, who knows, e.g. at one point e.g. newsfeeds.com apparently had been restructured so it is not unthinkable. i got reports by several UNS users that "no such article" bug connected to faulty message-id index resurfaced again, don't know if they fixed it.
I have that unchecked.do you have "resolve dns on every connect" in properties->general unchecked? if it is checked - uncheck it and try again.
I have tried it with the firewall disabled with no improvement.it might be your firewall as well.
Oh, I see... it only retries certain errors. Is that search criteria based on the text in purple in the console window?" I still have the default criteria, which doesn't include 'connect'.if it is not dns resolution, does it give connect error then after some time? how much time? (connect error is not retriable but if you added it to retriable errors yourself set retries to zero).
These are the errors I see:
"Error A connection attempt failed because the connected party did not properly respond after a period of time, or established connection failed because connected host has failed to respond. (System:10060)"
This one streams by a lot - I guess you're saying that's because it doesn't retry these, and just considers the header dead and skips it.
OR
"Error Select timeout error"
This I see more infrequently - and you're saying that's because it's doing my timeout, but also doing my retries - so 60sec * 30 retries, it will be 30 minutes before I see this error for a given header (if so configured).
Do I understand that correctly?
I dropped retries to 1 and timeout to 60 seconds, and now I'm getting many more of that timeout error. So I think I got it.
Well, I only signed up for a 3 day this time anyway with UNS. I wonder if there are other places that do that? With the Comcast limit, I don't need the full month - I could just setup a big queue and run it over the weekend once or twice a month.
Well, thanks for the feedback.