Installing PHP on RedHat ES4 with Apache 2

Apache 2 was installed during the system build of this RHEL ES4 system. Now, the task, to get MySQL 5.0 and PHP 5.2.3 installed. Here are some things I learned along the way.

First things first,
Remove any old MySQL versions with rpm -ev.
Also check for and remove any old php versions (rpm -qa | grep php)

Download the tar/gz bundle from php.net.

Download and install these packages from MySQL - I used community server:
MySQL-client-community-5.0.41-0.rhel4 (Client)
MySQL-shared-community-5.0.41-0.rhel4 (Shared Libraries - this may not be needed)
MySQL-server-community-5.0.41-0.rhel4 (Server)
MySQL-devel-community-5.0.41-0.rhel4 (Headers and Libraries)

Note that if you don't install the devel RPM you'll get this error on ./configure of PHP:
configure: error: Cannot find MySQL header files under yes.
Note that the MySQL client library is not bundled anymore!

It goes away after you install the devel RPM.

Then, it's time to ./configure PHP
I'm loosely following this doc: http://us.php.net/manual/en/install.unix.php
and Example 4.1 although I have Apache 2.0 instead of 1.3.
(the Apache 2.0 example gives some useful info also, but isn't strictly accurate either)

Prerequisites, per this page:
gcc
flex
bison

I installed these with up2date which may not give the latest versions, but hopefully good enough.

up2date gcc also installed:
cpp
glibc2
glibc-devel
glibc-headers
libgcc
glibc-common

I checked if PHP was currently loaded in the /etc/httpd/conf/httpd.conf file - looking for php in the "LoadModule" section. Not there. Keep going.

Noticed that in the example on php.net the ./configure specifies "with-apxs" so I wanted to see what apxs was and if I already had it.

What it is - tip of the keyboard to http://www.onlamp.com/pub/a/php/2001/03/15/php_admin.html
it's used to build and install Apache extension modules. After a few random searches, I found it's installed in the httpd-devel RPM. Of course it's not an apache-devel RPM, because RedHat prefers to disguise Apache with the httpd name. You tell me why, I'll give you a quarter.

up2date httpd-devel
and apxs magically appeared in the /usr/sbin directory.

httpd-devel said it depended on a bunch of other things too:
apr-devel
apr-util-devel
cyrus-sasl-devel
db4-devel
expat-devel
openldap-devel
pcre-devel
compat-openldap

First time I ran ./configure in the PHP directory (ignoring the example that says to install Apache first, since it's already installed) - I got this error:

./configure --with-mysql --with-apxs=/usr/sbin/apxs
...
checking for Apache 1.x module support via DSO through APXS... configure: error: You have enabled Apache 1.3 support while your server is Apache 2. Please use the appropiate switch --with-apxs2

Maybe we should tell the php folks how to spell "appropriate" except then I'd have to admit I didn't read the instructions first....oh well.

next run:
./configure --with-mysql --with-apxs2=/usr/sbin/apxs
next error:
checking for xml2-config path...
configure: error: xml2-config not found. Please check your libxml2 installation.

Oh where, oh where do I find xml2-config?
It's in libxml2-devel, of course! Now why didn't I know that???

up2date libxml2-devel
also installed this dependency:
zlib-devel

now this file exists: /usr/bin/xml2-config ...yay!checking for MySQL UNIX socket location... /var/lib/mysql/mysql.sock
configure: error: Cannot find MySQL header files under yes.
Note that the MySQL client library is not bundled anymore!

But if you already installed the MySQL-community-devel package (see note at beginning), you didn't get this right?

next run:
./configure --with-mysql --with-apxs2=/usr/sbin/apxs

got this error:
checking for gcc... gcc
checking whether the C++ compiler (gcc ) works... no
configure: error: installation or configuration problem: C++ compiler cannot create executables.

this web page gave me the next clue - to check the config.log in the directory where I'm doing the ./configure -
in the last few lines of config.log:
gcc: installation problem, cannot exec `cc1plus': No such file or directory

So now I'm on the hunt for cc1plus.
I found it through hints on random websites...it's in the gcc-c++ package.

up2date gcc-c++
which also installed:
libstdc++
libstdc++-devel

./configure --with-mysql --with-apxs2=/usr/sbin/apxs ... again
Oh my Dog...it actually worked:
+--------------------------------------------------------------------+
| License: |
| This software is subject to the PHP License, available in this |
| distribution in the file LICENSE. By continuing this installation |
| process, you are bound by the terms of this license agreement. |
| If you do not agree with the terms of this license, you must abort |
| the installation process at this point. |
+--------------------------------------------------------------------+

Thank you for using PHP.

Now...back to the php.net install on unix page.

make
which had a friendly reminder to run "make test"

make test

{sudo} make install
gave this reminder:
libtool: install: warning: remember to run `libtool --finish /home/julie/php-5.2.3/libs'

Since libtool doesn't exist on my system, so I installed it:

up2date libtool
which also installed
autoconf
automake

and running this:
libtool --finish /home/me/php-5.2.3/libs

said this:
PATH="$PATH:/sbin" ldconfig -n /home/julie/php-5.2.3/libs
----------------------------------------------------------------------
Libraries have been installed in:
/home/julie/php-5.2.3/libs

If you ever happen to want to link against installed libraries
in a given directory, LIBDIR, you must either use libtool, and
specify the full pathname of the library, or use the `-LLIBDIR'
flag during linking and do at least one of the following:
- add LIBDIR to the `LD_LIBRARY_PATH' environment variable
during execution
- add LIBDIR to the `LD_RUN_PATH' environment variable
during linking
- use the `-Wl,--rpath -Wl,LIBDIR' linker flag
- have your system administrator add LIBDIR to `/etc/ld.so.conf'

See any operating system documentation about shared libraries for
more information, such as the ld(1) and ld.so(8) manual pages.
----------------------------------------------------------------------

But otherwise php installed fine with a new module in /etc/httpd/conf/modules:
libphp5.so

and added this to /etc/httpd/conf/httpd.conf:
LoadModule php5_module /usr/lib/httpd/modules/libphp5.so

service httpd stop
service httpd start

And I should be good to go!

FSCK and a 4TB filesystem

It's not like I had a choice, the 4TB filesystem made of 10 x 500Gb SATA disks in RAID5 lost a disk and did a funky dance, and now it won't let anyone mount it except read only.

So I had to unmount it, with umount -l because when I tried plain old umount /name it said "device is busy". I don't know how it could be busy, I was the only person connected to the machine and no other server had it mounted...so I ran "umount -l /name" and that did the trick in about 20 seconds.

Then on to the fsck of the unmounted filesystem...did I mention it's 4 terabytes? What I would love to know...LOVE to know....is how long it takes to fsck a 4TB filesystem? A couple hours, a day, a weekend, or until eternity and the disks stop spinning?

I guess too if I had another wish it would be that there was some kind of moving blinking cursor to let me know fsck was doing something. After I said "yes" to a few dozen errors about block bitmaps, inode bitmaps, and inode tables with their ominous message "warning: severe data loss possible" then it spit out:

/dev/sdb contains a file system with errors, check forced.
Pass 1: Checking inodes, blocks, and sizes

where it has sat for the last hour, completely unresponsive.

Am I missing something? Did I miss the prompt for verbose output???

Answers:
- there is a prompt for verbose output but it didn't help much
- fsck on a 4TB disk took about 6 hours with lots of errors
- there is a progress bar using -C but I couldn't get it to display, probably missed something

Finally I ran it as root in the background, sent output to a log file, and went to sleep.

Disk partitioning, note to self

Next time I have to partition a *nix filesystem, I solemnly swear to read this post and adhere to the lessons.

This is what I've done recently, based on how the servers I've inherited were configured:

/ 512MB
/home 2Gb
/usr 2Gb
/boot 250MB
swap 2Gb
/var remaining space on 36GB disk.

The problems...running out of space on /home when I try to put a bunch of new rpms for RedHat that need to be installed. That /opt, which some apps like to use as the default path, is part of / and at a measly 512MB it goes too quickly.

This is what I plan to do next time...

/ 1GB
/home 4GB
/usr 4GB
/boot 500MB
swap 2 -4 GB dep on physical memory
/var less of the 36 GB remaining, but still a lot for mostly log files

APC Smart UPS and Red Hat Linux

I am forever amazed at how documentation on how to install products on Linux which were primarily made for Windows is sufficiently lacking in every respect. Okay, I'm not that amazed. But I came into the Unix world through the back door - I started with Novell NetWare 3.x and Windows for Workgroups 3.51 (remember those!) and then meandered through the world of Windows NT Server and found myself in the networking world of Cisco routers and such.

Then, a few years later, I met Unix. I can't say it was love at first sight, in fact it was more like chaos and insanity, but now, Unix and I, we understand each other.

So today I was figuring out how to install the server agent on the APC Smart UPS SC 1500 I just bought. It's a RHEL ES 3 server with a serial interface, so I connected it and didn't expect any sort of miracle that the server would know it had something new attached. It didn't, of course.

Unix agent, I told myself, I need to find the Unix agent. I found this in the instructions on how to install on Unix - basically I needed to find the RPM.

Now, of course, that I'm looking for the info I found a handy text doc on how to do this:

RPM is on the cdrom.
rpm -i pbeagent-7.0.4-114.i386.rpm
will get the agent installed.

The install puts a file PBEAgent in /etc/init.d - this is what I used to figure out where the installation was.

Agent installs to /opt/APC/PowerChuteBusinessEdition (of course! that's so obvious!)
then by chance I found this file:
config.sh
which I thought I'd run just to see what happens.

Turns out, this is a good thing, because I chose "simple" for the install and it prompted me to create username/password, select the com port (I guessed it was com1).

Then, I wondered, what next?

Ah, install the Windows server software. Somehow, autotragically, the server s/w install discovered my server running the powerchute agent. Somehow it's all working, and as with a lot of windows originated software, how is still a mystery.