Open GoogleCodeExporter opened 9 years ago
上面最后一行单位是写错了,是jap能达到1MB/s的速度,也就是
撑满我的带宽
Original comment by fifa...@gmail.com
on 10 Mar 2013 at 4:16
0.21.0
将Range机制独立出来抽象成一个api了,并复用到了C4上(还有�
��虑复用到Forward上),可能还有一些隐藏的问题待查。
snova/gsnova的proxy是HTTP
Proxy,相比socks方式在代理https上多了一次HTTP
CONNECt连接协商过程,所以在连接过程中相比jap有延迟。
相比JAP的速度问题可能和额外的HTTP解析以及TCP连接复用有关�
��snova限制全局到C4服务端只有MaxConn个长连接,
JAP则是每个Proxy创建一个连接)。
Original comment by yinqiwen@gmail.com
on 10 Mar 2013 at 5:46
我把服务端从go换回了java的,貌似也无任何改善。
还有新版本出现了大量这种log,是什么原因?
2013/03/10 14:50:27 c4_http.go:141: Pull worker:4 stop working
2013/03/10 14:50:27 c4_http.go:128: Pull worker:4 start working
2013/03/10 14:50:29 c4_http.go:141: Pull worker:1 stop working
2013/03/10 14:50:29 c4_http.go:128: Pull worker:1 start working
2013/03/10 14:50:29 c4_http.go:141: Pull worker:12 stop working
2013/03/10 14:50:29 c4_http.go:128: Pull worker:12 start working
2013/03/10 14:50:34 c4_http.go:141: Pull worker:13 stop working
2013/03/10 14:50:34 c4_http.go:128: Pull worker:13 start working
2013/03/10 14:50:37 c4_http.go:141: Pull worker:4 stop working
2013/03/10 14:50:37 c4_http.go:128: Pull worker:4 start working
2013/03/10 14:50:39 c4_http.go:141: Pull worker:12 stop working
2013/03/10 14:50:39 c4_http.go:128: Pull worker:12 start working
2013/03/10 14:50:44 c4_http.go:141: Pull worker:10 stop working
2013/03/10 14:50:44 c4_http.go:128: Pull worker:10 start working
2013/03/10 14:50:44 c4_http.go:141: Pull worker:13 stop working
2013/03/10 14:50:44 c4_http.go:128: Pull worker:13 start working
2013/03/10 14:50:48 c4_http.go:141: Pull worker:4 stop working
2013/03/10 14:50:48 c4_http.go:128: Pull worker:4 start working
2013/03/10 14:50:49 c4_http.go:141: Pull worker:12 stop working
2013/03/10 14:50:49 c4_http.go:128: Pull worker:12 start working
2013/03/10 14:50:54 c4_http.go:141: Pull worker:10 stop working
2013/03/10 14:50:54 c4_http.go:128: Pull worker:10 start working
2013/03/10 14:50:55 c4_http.go:141: Pull worker:13 stop working
2013/03/10 14:50:55 c4_http.go:128: Pull worker:13 start working
2013/03/10 14:50:58 c4_http.go:141: Pull worker:4 stop working
2013/03/10 14:50:58 c4_http.go:128: Pull worker:4 start working
2013/03/10 14:51:00 c4_http.go:141: Pull worker:12 stop working
2013/03/10 14:51:00 c4_http.go:128: Pull worker:12 start working
2013/03/10 14:51:04 c4_http.go:141: Pull worker:10 stop working
2013/03/10 14:51:04 c4_http.go:128: Pull worker:10 start working
2013/03/10 14:51:05 c4_http.go:141: Pull worker:13 stop working
2013/03/10 14:51:05 c4_http.go:128: Pull worker:13 start working
Original comment by fifa...@gmail.com
on 10 Mar 2013 at 6:53
你把MaxConn改的太大了。这个是保持到C4的长连接,每个连接��
�多保持ReadTimeout秒(这个又设置的太短)。
注意这个在不支持HTTP chunked
response的PaaS上要将ReadTimeout改为3才正常工作(如Appfog)
Original comment by yinqiwen@gmail.com
on 10 Mar 2013 at 6:56
readtomeout 我用的新版默认的 200,
貌似上一版默认是25,这个单位是s还是ms……
maxconn我用了15……,设置了3个node,
maxconn这一版默认是1,也太小了吧?
Original comment by fifa...@gmail.com
on 10 Mar 2013 at 7:11
单位秒。
没做过测试。在网络比较好的情况下,设置为1工作地很正常�
��没有太明显的延迟情况。(不过设置为1会大大降低PaaS服务�
��负载)
Original comment by yinqiwen@gmail.com
on 10 Mar 2013 at 7:15
我开了3个C4
node,如设置maxconn为1的话,只有一个工作,其他两个一直歇��
�?
另外一直不是很了解C4设置多个node有无帮助?
Original comment by fifa...@gmail.com
on 10 Mar 2013 at 7:21
都会工作的,顺序选择的方式。
第一个Proxy连接的请求选择第一个C4
Node,第二个选择第二个node,以此类推
Original comment by yinqiwen@gmail.com
on 10 Mar 2013 at 7:23
顺便问下:下面这个log怎么会出现的,我设置了自定义的googl
eIP,里面并没有74.125
.128.86这个ip,我看cloud
host里面也没这个ip,但是启动gsnova时为何要去连这个ip呢?
2013/03/10 15:22:06 gae.go:582: Init GAE.
2013/03/10 15:22:08 gae.go:260: Failed to request data from GAE:xxxx.apps
pot.com for:http: error connecting to proxy https://GoogleHttps: dial tcp 74.125
.128.86:443: i/o timeout
2013/03/10 15:22:08 gae.go:190: http: error connecting to proxy https://GoogleHt
tps: dial tcp 74.125.128.86:443: i/o timeout
2013/03/10 15:22:08 gae.go:627: Failed first to auth appid:http: error connectin
g to proxy https://GoogleHttps: dial tcp 74.125.128.86:443: i/o timeout
Original comment by fifa...@gmail.com
on 10 Mar 2013 at 7:28
顺便问下,我设置了自定义googleip,里面并没有某个google ip
比如:74.125.128.86
我看cloud host里也没这个ip,
但是日志里有时却会出现类似dial tcp 74.125.128.86:443 time out
云云,这是怎么回事
btw: 怎么今天连发几条log都被系统吃了……
Original comment by fifa...@gmail.com
on 10 Mar 2013 at 7:41
如果[GAE]下的Proxy=有设置,默认连接的是这个地址;同时这个
地址又可以在cloud_hosts.conf和user_hosts.conf中配置(user_hosts优先
)
Original comment by yinqiwen@gmail.com
on 10 Mar 2013 at 7:46
[GAE]下有Proxy=https://GoogleHttps,就是说只要有这个在就是连接74
.125.128.86这个固定的ip?
可能我没有表达清楚我的意思,我已经在user_hosts.conf中设置��
�自定义的googlehttps ip,但一上来连GAE没有根据我的这些ip来
我希望的结果是凡是要连google的443端口,统一从我的user_hosts.
conf里面的列表里面尝试,不管是一上来连GAE还是通过根据spac
匹配访问google服务的,该如何设置?谢谢~
Original comment by fifa...@gmail.com
on 10 Mar 2013 at 7:57
你的user_hosts.conf怎么配置的?
Original comment by yinqiwen@gmail.com
on 10 Mar 2013 at 8:02
gsnova.conf
[Google]
Enable=1
ConnectTimeout=1500
PreferIP=true
Proxy=
user_host.conf
#myappid.appspot.com=1.1.1.1
GoogleHttpsIP=*.*.*.*|*.*.*.*|*.*.*.*|*.*.*.*|*.*.*.*
Original comment by fifa...@gmail.com
on 10 Mar 2013 at 8:08
相应的[GAE]下Proxy改为Proxy=https://GoogleHttpsIP
[Google]下PreferIP=true
Original comment by yinqiwen@gmail.com
on 10 Mar 2013 at 8:18
OK,
thanks
Original comment by fifa...@gmail.com
on 10 Mar 2013 at 8:34
经过测试,这个issue也fix了
Original comment by fifa...@gmail.com
on 10 Mar 2013 at 4:25
Original issue reported on code.google.com by
fifa...@gmail.com
on 10 Mar 2013 at 3:52