[ale] A new web server

Jim Kinney jim.kinney at gmail.com
Mon Jan 19 22:57:10 EST 2009


On Mon, Jan 19, 2009 at 10:41 PM, Terry Bailey <terry at bitlinx.com> wrote:
>
>
>
>
>
> The random errors are random within a particular subcollection of
> items.  I went to a computer store today in Douglasville and the same
> thing happened on one of their XP computers as happens on the XP
> machine that I am running.   So, when I got home, I whipped out a
> laptop and hooked up to the wireless on my home network.  It is an XP
> machine and works perfectly time after time and I flush the browser
> cache after each trial.
>
> So the server is performing well at my house for the Fedora 9 box and
> the XP laptop, just not my XP tower nor the XP tower in Douglasville.
>
> I think that it may be a flaky router because I am in a boutique data
> center and the ownership has just changed hands.
>
> Since I am running CentOS 5.2 64 bit on the new server and since my
> take is that this is just Red Hat Enterpirse Linux, it seems that the
> OS ought to be OK.  I downloaded via BitTorrent and burned the DVD
> myself.  Maybe there is a problem there.

The 64-bit CentOS 5.2 is basically rock solid. Especially as a web
server. A bad burn of the DVD will usually show up in the install not
later during use.

You can test the binary checksums with rpm -Va to verify that all
installed code is what it should be according to the installed record
in the rpm database.
>
> Memtest is not on the server, but I will try to find it and I am not
> sure how to do a tcpdump.

Memtest is run from a bootable CD/DVD. It's on the install disk as a
boot option.

tcpdump is run as root. You will need to spec the NIC to listn to as
well as the port. It's best to dump to a file for a period of time.
man tcpdump will give _all_ the gory details. Reading the output may
cause blindness so polish up the grep skills.

tcpdump -i eth1 -n -p port 80 -c 1000   will capture 1000 packets on
NIC eth1 only using port 80 and will not do a DNS lookup nor put the
NIC in promiscuous mode.
>
>
>
> At 10:05 PM 1/19/2009, you wrote:
>>On Mon, Jan 19, 2009 at 4:50 PM, Terry Bailey <terry at bitlinx.com> wrote:
>>
>> > BTW, this problem is intermittent.
>> >
>>Download the 2.6.28 kernel and compile specifically for your machine.
>>There are numerous "gotcha's" in the 2.6.27 kernels in F10 at the
>>moment. I also suspect a flaky router at an ISP. Run tcpdump on the
>>web port on both ends of the connection (server and client). If the
>>transfer byte count differs, call and send the logs to the ISP network
>>engineering team for both the server and client ends.
>>
>>Also test the same page multiple times and flush the browser cache
>>after each attempt. If the missing parts are inconsistent, that
>>further suggests router problems.
>>
>>Last test: run memtest on the server. I can't count the number of
>>problems that go away when the system passes all memtest trials. Ram
>>just a tad flaky is OK for clients but a nightmare to debug on
>>servers. The only thing memtest can't do is run the hard drives full
>>steam during the ram tests. This will also stress test the power
>>supply than can also cause intermittent ram issues.
>>
>>--
>>--
>>James P. Kinney III
>>_______________________________________________
>>Ale mailing list
>>Ale at ale.org
>>http://mail.ale.org/mailman/listinfo/ale
>
> _______________________________________________
> Ale mailing list
> Ale at ale.org
> http://mail.ale.org/mailman/listinfo/ale
>



-- 
-- 
James P. Kinney III


More information about the Ale mailing list