Hello system admins,
Here is my usual very late summary. It took me a long time to get to the
bottom of this problem. It turns out that there was nothing wrong with
any of the apache startup scripts I attempted (and I tried many,
including the "apachectl" script that comes with the distribution). Nor
was there a problem of where I placed the startup script -- /sbin/init.d
or elsewhere. Also, it was not necessary that I put a "nohup" in front
of the command that actually starts apache. Thanks to all who responded
with these and other related suggestions, even if they didn't fix my
problem. I appreciate your input anyway, and I learned much from it.
The problem was that my AdvFS root_domain was corrupted, but not enough
that it would cause any really obvious problems. Working with Compaq
support, I managed to run the command "/sbin/advfs/verify root_domain"
in the correct way (the man page was not clear on this). This showed
errors indicating corruption. According to support, the only way to run
verify on the root domain AND know that any errors that show up are
significant, is to boot off the distribution CD, use /dev/MAKEDEV to
create the system device, manually re-create the /etc/fdmns directory
structure that has the link to the system device, AND THEN run the verify
command.
So that's what I did to actually show that the internal AdvFS file
structure had some corruption. Using the "verify -f" command couldn't
completely rid the disk of the problem, although it did do some good:
After using the "-f" switch, my apachectl startup script in /sbin/init.d
began working on every boot. What a relief!
To completely rid the system disk of the corrupt AdvFS, I had to re-create
the root domain and fileset on that disk, and then restore from a tape
backup (vdump). (I did not have to re-label using disklabel.) More
specifically, the steps were:
Obtain a good backup of the root fileset using vdump
In my case, I booted single-user, loaded a new tape, then ran
# /sbin/vdump -0 -F 64 /
to get a backup. Note that I had only root and swap on
that entire disk - no /var or /usr, etc.
Boot off the Tru64 v4.0F media CD
Invoke the shell command prompt
# cd /dev
# ./MAKEDEV re5 <---- my root device
# ./MAKEDEV tz13 <---- my tape device
# mkfdmn /dev/re5a root_domain
You will have to confirm that you want to replace (destroy) the
domain that currently exists on that partition.
# mkfset root_domain root
# mount root_domain#root /mnt
# cd /mnt
Make sure the tape-to-restore-from is ready
# vrestore -xv ./
# cd /
# umount /mnt
# /sbin/advfs/verify root_domain
This last command gave normal output with no errors (it had to be
error-free since the domain was just created). Finally, it is
free of corruption.
# shutdown -h now
# boot <---- Now you can boot normally, all done.
My original post is at the end. Thanks very much to the following for
taking the time to respond to my question:
Jeff Berliner <jeff_at_endeavor.med.nyu.edu>
"Sloane, Robert Raymond" <sloane_at_ukans.edu>
Frank Wortner <frank_at_bondnet.com>
Daniel Monjar <dam_at_orgtek.com>
Bill.Melvin_at_esc.edu
mes Sainsbury <sainsb_j_at_chem.usyd.edu.au>
Nikola Milutinovic <Nikola.Milutinovic_at_ev.co.yu>
Ivan Hoe <ivanhoe_jack_at_hotmail.com>
Regards,
Paul Youngblood
---------------------------------------------------------------------------
------------ Original Posting
---------------------------------------------
---------------------------------------------------------------------------
Date: Tue, 27 Feb 2001 16:04:18 -0600 (CST)
From: Paul N. Youngblood <youngbp_at_uahis1.uah.edu>
To: tru64-unix-managers <tru64-unix-managers_at_ornl.gov>
Subject: Unable to start apache server at system boot
Hello Sys Admins,
I've got Tru64 v4.0F with Enhanced Security and Apache v1.3.13.
I can start the server interactively when logged on as user
"root" -- either by invoking the daemon directly like this:
# /usr/local/apache/httpd
or, by invoking it with a script like this:
# /sbin/init.d/apachectl start
Once running, the server functions without problem or error.
However, it never starts during a system boot. (If it actually
is started, it must die immediately.) Nothing ever gets logged
about apache during bootup -- error or otherwise in the system
log files, /var/adm/syslog.dated/current/kern.log and
/var/adm/syslog.dated/current/daemon.log
And no entry is ever made to either of these apache log files:
/usr/local/apache/logs/error_log
/usr/local/apache/logs/httpd.pid
Of course, when I start apache by hand, the httpd.pid file is
overwritten with the current pid, and a line like the following
is written to the error_log file:
[Mon Feb 26 19:13:45 2001] [notice] Apache/1.3.14 (Unix)
configured -- resuming normal operations
My startup link looks like this:
# ls -l /sbin/rc3.d/S99apache
lrwxrwxrwx 1 root bin 19 Feb 26 16:34
/sbin/rc3.d/S99apache -> ../init.d/apachectl
My startup script permissions/owners look like this:
# ls -l /sbin/init.d/apachectl
-rwxr-x--- 1 bin bin 7121 Feb 26 19:13 /sbin/init.d/apachectl
Any and all suggestions/ideas appreciated. Thanks.
Paul Youngblood
youngbp_at_uahis1.uah.edu
----------------------
Received on Mon Apr 16 2001 - 22:57:32 NZST