zrlw

profile填加了邮箱，就是同名sina邮箱，欢迎交流。 我们现在压测环境还触发了一个issue还没有定位： #7698， 压测运行一两天莫名其妙有一个节点提示I/O Reactor STOPPED，节点同步给其他节点的httpclient线程挂了，后果就是这个节点负责的实例无论增加还是减少都是只有它自己知道，其他节点一概收不到这个节点的同步请求。

> 首页README.md 有钉钉群二维码

> If execute report to all nodes each time, whether will cause high pressure? The report only interval 2 seconds. it's a trade-off solution. let's say there are 10 nodes...

我们要用nacos sync做多机房实例同步，用不了2.0，因为2.0目前的机制和nacos sync不兼容。如果自己定制的版本还有问题，就打算换软件了。

>  去年我问过这个问题，nacos官方回应短期内不会考虑。 > > 如果你们可以实现一致性hash方案，能否提个PR？ https://blog.csdn.net/github_28583061/article/details/106384335 pr不打算提了，因为我们的方案是结合前端负载均衡一起做一致性hash，并不通用

我在nacos社区群2，钉钉搜不到你啊

internal-sftp reads up to 100 entries in one message. https://github.com/openssh/openssh-portable/blob/730a80609472ee0451c99482d75c9c41f3ebc42d/sftp-server.c#L1161 sftp client will keep sending READDIR request until receiving SSH_FX_EOF signal from sftp server, such as: https://github.com/is/jsch/blob/addb8e3a0ebd71cbbf51cf6ba746e8e128df713c/src/main/java/com/jcraft/jsch/ChannelSftp.java#L1639 it might be...

The default mod_sftp channel packet size (channelPacketSize) of proftpd is 32KB, it could only take about 20+ entries for each readdir operation.

the default value of ```FSCachePolicy``` (internal filesystem-related cache) should be set to ```off``` instead of ```maximum number of entries is 3000, and the default maximum age is 3 seconds```. http://www.proftpd.org/docs/modules/mod_core.html#FSCachePolicy...

> @zrlw Would you care to verify that #1541 does what you'd prefer, in your testing/setup? it works for us.