> Dnia wtorek, 29 listopada 2011 20:48:52 Roberto De Ioris pisze:
>> I have committed another series of fix/optimizations.
>>
>> - nodes marked as dead will be skipped even if their reference is > 0
>> (only already accepted request will continue)
>> - added another check in auto-opmtized-list consistence (i suppose this
>> is
>> the one throwing your error)
>> - a potential segmentation fault in slot removal should be fixed
>>
>> Let me know if it definitely fix it for you
>
> I've noticed one more possible issue so far, requests field for one node
> gets
> zeroed but that should only happen if I restart fastrouter. Below is part
> of
> two subsequent reads from fastrouter stats socket.
>
> 1. first fastrouter stats output
> "hits": 433088,
> "nodes": [
> {"name": "172.16.200.55:3001", "modifier1": 0,
> "modifier2": 0, "last_check": 1322681099, "requests": 12679, "tx":
> 104686291,
> "ref": 1, "death_mark": 0},
> {"name": "172.16.200.56:3001", "modifier1": 0,
> "modifier2": 0, "last_check": 1322681099, "requests": 12489, "tx":
> 103283916,
> "ref": 0, "death_mark": 0}
> ]
>
> 2. second fastrouter stats output
> "hits": 433089,
> "nodes": [
> {"name": "172.16.200.56:3001", "modifier1": 0,
> "modifier2": 0, "last_check": 1322681100, "requests": 12490, "tx":
> 103283916,
> "ref": 1, "death_mark": 0},
> {"name": "172.16.200.55:3001", "modifier1": 0,
> "modifier2": 0, "last_check": 1322681100, "requests": 0, "tx": 0, "ref":
> 0,
> "death_mark": 0}
> ]
>
>
When nodes got unsubscribed, they are completely removed. Think about
using the form:
socket = 192.168.0.*:0
where the same instance could bind to random ports (for example i now
configure most of my instances in this way), you will start to fill the
linked-list at each restart.
This is obviously not your case, but the less elements the linked list
has, faster lookup you will get. I suppose some form of "history" could be
added in the future.
A short-term solution would be adding a "nail" option like
--nail-subscribe-to <server> where "nailed" nodes remains in the list
forever and non-nailed nodes (the default) continue to behave like before.
The code would be as simple as
(from)
if (node->death_mark && node->reference == 0) {
remove_node(node);
}
(to)
if (node->death_mark && node->reference == 0 && !node->nailed) {
remove_node(node);
}
What about it ?
--
Roberto De Ioris
http://unbit.it
_______________________________________________
uWSGI mailing list
[email protected]
http://lists.unbit.it/cgi-bin/mailman/listinfo/uwsgi
