Server

访问听下包发现：

两个（组）前端：fronta和frontb，跑 Linux + Squid。图片（img）、动态页面采用同一组cache，后端使用apache 2.2.6，php 5.2.4。所有图片、静态js均假定用户不刷（i.e. 设置一个较长的过期时间）来减轻负载。

Mars is a Planet (slightly modified) installation based RSS reader, which has built my frequently visited feeds into it. The domain name is proposed by Neo Gui who thinks that ‘planet’ would be better off the Mars.

What I liked is that Planet, unlike many of other RSS readers, renders a static page for user consumption. This makes it to be more reliable when the server itself is handling dynamic contents, and much less chance being an attack surface.

据 eWeek 说，OpenSolaris快要有IBM Mainframe上跑的版本了。看来，这个市场已经快要萎缩到昔日的竞争对手开始同仇敌忾的地步了。

Complains about can not reach www.freebsd.org was popped up on FreeBSD mailing list. This is believed to be caused by hardware issue, as reported by Simon Nielson recently.

去掉以后，下列症状消失：

pid 11673 (httpd), uid 80: exited on signal 11
pid 25946 (httpd), uid 80: exited on signal 11
pid 26292 (httpd), uid 80: exited on signal 11
pid 13623 (httpd), uid 80: exited on signal 11
pid 27096 (httpd), uid 80: exited on signal 11
pid 31838 (httpd), uid 80: exited on signal 11
pid 30847 (httpd), uid 80: exited on signal 11
pid 38579 (httpd), uid 80: exited on signal 11
pid 38269 (httpd), uid 80: exited on signal 11
pid 34011 (httpd), uid 80: exited on signal 11
pid 38698 (httpd), uid 80: exited on signal 11
pid 32137 (httpd), uid 80: exited on signal 11
pid 38754 (httpd), uid 80: exited on signal 11
pid 38756 (httpd), uid 80: exited on signal 11
pid 38755 (httpd), uid 80: exited on signal 11
pid 38759 (httpd), uid 80: exited on signal 11
pid 38760 (httpd), uid 80: exited on signal 11
pid 39259 (httpd), uid 80: exited on signal 11
pid 39253 (httpd), uid 80: exited on signal 11

呵呵，在互联网公司工作的时间长了，会有很多术语。比如产品的「上线」，其实对很多人来说就会是一件很有压力的事情，所以有些时候大家会避开用这个词。

如许多人了解的那样，ftp.freebsdchina.org的服务器挂了，很不幸地，这件事多少也有我一份，在硬件故障之后，我怀疑我尝试修复的最后一次操作给了它致命一击——这台在机房运行了四年之久的机器，终于消失在了机房茫茫闪烁的指示灯中，更为严重的是，我们恐怕永远无法再找到这台机器，因为没有人知道它在机房的确切位置了……

散乱的记下来备忘。看之前留神，如果被我搞糊涂了可能就彻底糊涂下去了……

一些需求：
a) 满足尽可能多的集群节点的更新需要，高承载能力。
b) 失败时能够迅速完成回退。
c) 确保安全，这包括数据完整性、访问控制等。
d) 部署尽可能的自动化。
e) 最小化downtime。

一些潜在的先决条件：

I have finally got some progress of the Bayesian traning on tarsier. Now it has learnt the data from the past six months, both wrongly detected hams and spams. I hope this can improve the accurancy of the anti-spam system.

Peter早上发信说这台机器(committer转信以及邮件列表等)上许多芯片烧掉了，还有一些芯片上出现了穿孔并露出了硅片……不过他迅速地更换了硬件。由于FreeBSD.org使用了多个MX，因此这一故障并未导致邮件丢失。