Auto-retrying issue

[Greg_G]

Even though this may be related to the SSL timeout issue (now resolved), it even more likely is due to a misconfiguration on my part. So I decided to start a fresh thread. While the SSL timing out is now working wonderfully, I'm continuing to have trouble with the automatic retrying. I suspect I just have an incorrect setting, however. Here's what I have:

retries: 5
retry failed article tasks every: 60 min
recoverable errors:too*many|limit|timeout|ssl*read|ssl*connect|actively*refused|properly*respond

Despite this, it seems that UE gives up after the first attempt and never automatically retries. If I manually force a retry it does so.

Any ideas what I've got set wrong? I think it's only showing up so badly because of the severe UNS SSL issues. But that makes it a good time for me to try to figure out. Thanks.

[alex]

what is the error? it is in the error red question sign task manager pane.

all SSL errors are retriable, maybe 5 retries is not enough?
if keep alive is enabled you can try to add connection delay.

[Greg_G]

I'm getting various errors, all the ones that have been plaguing me on SSL: SSL read error, SSL connect error, actively refused, et al. UNS SSL is just a mess for me right now. (I'll contact UNS support about it once I get these kinks worked out.)

Interestingly (and I obviously forgot to mention this) they do not pass into the red question mark tab (headers do, but not articles). Yet, after only the one attempt, they show as "secure.usenetserver.com retries: exhausted/error" in properties and don't seem to retry after an hour.

I'll go note down a few tasks and make 100% sure they haven't retried after 60 minutes. But I'm pretty sure they are not doing so. I did that yesterday with 1.9.9.3 and they didn't, but a double-check doesn't hurt.

[alex]

no, exhausted means exhausted, all works ok.

just if it is partial and one part exhausted the other parts will still be tried before the partial goes to the error pane. you can see it only in properties since i felt it will be too much to add V ? interposed state icon, so it is just V until all parts are done then you'll see ? on the article and it will go to the error tab.

you may try to increase the number of retries and add connection delay, just few seconds (properties->servers), otherwise UE may bombard UNS with frequent connection attempts and maybe the server is not in shape to take that (SSL connection is much more heavier to estabilish, with keep alive enabled it is always full handshake).

probably more users are downloading through SSL so their server cannot handle it properly, i wonder what else it could be.

[Greg_G]

Okay Alex, thanks. I will increase the retries and the connection delay.

However, I just want to add one more piece of information on this. For all this SSL testing I had set the tasks value to 0 for the non-SSL server. Normally I keep the non-SSL server set to 2 tasks but with a strict priority of Low while the SSL server has a strict priority of Highest.

Even though I just restarted UE a little while ago to use 1.9.9.5, if I give the non-SSL server a tasks of 2, it will immediately start working on the articles that previously failed just once on the SSL server. Obviously it hasn't been 60 minutes * 5 retries yet.

Shouldn't it wait until all the retries on the SSL server have failed before trying the lower priority server? Or is one failure enough for it to switch to it? I'm hoping it's the former which makes more sense to me.

[alex]

only 5 retries matter, after exhausted it will try a lower preference server.

it doesn't have to go to the error pane in between and sit there for the retry period first.

UNS is acting strange lately.

[Greg_G]

only 5 retries matter, after exhausted it will try a lower preference server.

But, unless I'm misunderstanding, that's not what is happening. Even though I have it set to 5, it's goes to the next lower preference server after only a single failure (either that or it is trying 5 times immediately without waiting 60 minutes in between). Wouldn't this indicate a problem?

UNS is acting strange lately.

That's for sure!

[alex]

yes, misunderstood.

retry failed article tasks option in properties->tasks means once in a while it retries failed tasks from the error pane.

the number of retries in properties->servers refers to the number of retries (can be shortly one after another or if you set connection delay in properties->servers - after connection delay) before it lands in the error pane - unless there are other parts which didn't exhaust all retries within the same partial message.

when retries failed articles kicks in all retries start all over again so if you have 5 retries and 60 minutes retry failed it will make 5 retries, then rest for 60 minutes then makes 5 retries again etc.

[Greg_G]

Sorry for not getting back before now, I have a bad flu.

So, from what you are saying, there's no way to make UE stick with a top priority server for, say, 5 total recoverable tries, one every 30 minutes, over 2.5 hours, and only after failing all of those then try the next lower server?

[alex]

no, if it fails after exhausting all retries or after encountering non-retriable error it will go to a lower preference server.

you can detach secondary server and attach it when you really need it.

i had just sticked to the newspro options when adding this part so newspro users can make a transition, if not newspro it would be even less, generally people get perplexed with too many options.

also times are different, mostly what we have now popular are unlimited accounts so it doesn't matter which server is has higher preference.

[Greg_G]

Aye, okay Alex, thanks for all the info. Hopefully UNS will get their SSL act together and it'll be a moot point soon.