On Tue, Nov 5, 2013 at 4:53 PM, Pavel Levshin <[email protected]> wrote:
> > 05.11.2013 19:26, Rainer Gerhards: > > > Because this again requires code to be maintained for a border case. As I >> said, I wouldn't wonder if you get a segfault under some configurations >> where this is set to one (e.g. all watermarks are 0 in this case). >> > > Nobody has complained yet, right? yeah, and if someone complains, we probably spend weeks working on a border case... > Maybe it will break, maybe not. If the problem is in watermarks, then it > is not a problem of queue size. > > Btw, I've just tested it - and it works. Have you really tested it? with pure disk queues, DA queues, linkedList and FixedArray? . > > > I *really* don't see any valid use case for a queue with size 1... What is >> it? We were just about direct mode. Size 1, assuming it works, is >> essentially direct mode with a lot of overhead. So use direct in this >> case. >> Does that sound like an undue restriction? >> > > No, it is not the same as direct queue. In async queue, you can drop > messages when the queue is full (even if it's size is 1). In direct mode, > you must process all the messages. > > As of today, I do not have a real use case for this configuration. But it > can be used, for example, to deliver *some* of messages to a very slow > output with minimum delay. You can even deliver a sample of message each 1 > second. > > well.. strange use cases may pop up every now and then. The question is what is to best benefit of the project. Take the current situation. We started with global variables, something that has broader use, but still is not a mainstream use. That implementation lead the path to an engine rewrite. Again, this is not something that is purely done for that reason, but it wouldn't have been done if we hadn't started with that strange use case. I still think its good to do, and I hope the outcome will benefit everyone. But we need to see the price we need to pay: much more demanded features (e.g. the dynafile enhancements, lookup tables, ...) have been deferred and free support from my side has come to a standstill. That rewrite has essentially put on hold most of the project's activities. While I hope that the majority of things can be done by end of month, it will affect project agility up into the next year. *This* is why I try not to get too many border cases into the code... Rainer > > -- > Pavel Levshin > > _______________________________________________ > rsyslog mailing list > http://lists.adiscon.net/mailman/listinfo/rsyslog > http://www.rsyslog.com/professional-services/ > What's up with rsyslog? Follow https://twitter.com/rgerhards > NOTE WELL: This is a PUBLIC mailing list, posts are ARCHIVED by a myriad > of sites beyond our control. PLEASE UNSUBSCRIBE and DO NOT POST if you > DON'T LIKE THAT. > _______________________________________________ rsyslog mailing list http://lists.adiscon.net/mailman/listinfo/rsyslog http://www.rsyslog.com/professional-services/ What's up with rsyslog? Follow https://twitter.com/rgerhards NOTE WELL: This is a PUBLIC mailing list, posts are ARCHIVED by a myriad of sites beyond our control. PLEASE UNSUBSCRIBE and DO NOT POST if you DON'T LIKE THAT.
