Forum OpenACS Development: Installing Oracle 8.1.7 on Redhat 8.0

I installed oracle 8.1.7 on Redhat 8.0, and everything seems to work fine.    My plan was to update the install docs with a note about using the same install steps for redhat 8.0 and redhat 7.3, but I ran into a problem with installing openacs, and I wanted to check and see if anybody else has encountered the same issue.

While installing openacs, the install process would hang during the installatiion of the content-repository sqlj and jar files.  Top showed a jre running at 100%, and after about 10 minutes, I killed aolserver and aborted the installation.  I also stopped oracle and restarted it and the listener process.  While doing the initial installation of oracle, I saw this same behavior when trying to execute runinstaller.  Thinking that this problem might the same problem I was seeing with the oracle installer, I used the same work-around and added the sourcing of the  /usr/i386-glib21-linux/bin/i386-glibc21-linux-env.sh script and the setting of the LD_ASSUME_KERNEL=2.2.5 environment variable in the oracle startup and listener scripts.  And for good measure, I also added these in the nsd-oracle script that I was using to start aolserver.  After I did all this and restarted everything and restarted the openacs installation everything went fine.  Has anybody else had to do this for redhat 7.3 or 8.0?  I'm not sure all of these changes are necessary, but I'm too lazy to try re-installing  several times in order to isolate what is going on.

Collapse
Posted by Janine Ohmer on
It's the LD_ASSUME_KERNEL that fixes this, at least that was my experience.

Caveat - I haven't installed on Redhat 8 yet but I had this same problem on 7.2/7.3 (I forget which it was).  Also I don't use the compat libs - you can do it without them if you have the glibc patch from Oracle.  I think the end result is the same.

Collapse
Posted by Dan Wickstrom on
Where did you add it?  Di d you have to add it in the oracle and listener  scripts , and the nsd startup script as well, or in some subset of those three scripts?

On an unrelated note.  The notifications for these new forums seem to be unreliable.

Collapse
Posted by Jeff Davis on
on unreliable forum notifications:
I don't think I have missed any.

Do you have a particular post you did not get?

I can look in the server logs and see if there
are any errors that might explain why you missed
a notification.

Collapse
Posted by Dan Wickstrom on
Yes, it was Janine's response to my original post.  I just happened to still have the browser  open to the page where I did my original submission, and by chance I did a reload and saw Janiine's response.  I then checked my inbox, and noticed that I haven't received a notification yet.  I've noticed this a few other times too for other posts, but I didn't say anything at the time, because I thought it might have been startup pains for the new site.
Collapse
Posted by Jeff Davis on
I got that one via email but I will look in the logs to see if
there was an error processing the notifications list or
if the server restarted or something.
Collapse
Posted by Jeff Davis on
Dan, Roberto checked the qmail logs and found
@400000003dd92792096ff654 info msg 359911: bytes 1036 from <notification-62380-2960@openacs.org> qp 25381 uid 7791
@400000003dd92792099c51cc starting delivery 1885843: msg 359911 to remote danw@rtp.ericsson.se

@400000003dd927930cabf714 delivery 1885843: success: 198.24.6.3_accepted_message./Remote_host_said:_250_2.0.0_gAIIBtW11291_Message_accepted_for_delivery/
so it looks like the problem is on your side and not with notifications.
Collapse
Posted by Dan Wickstrom on
My apologies.  Sorry for the inconvenience.
Collapse
Posted by Jeff Davis on
Not a problem at all, I think it's important to hear about these kinds of things since it was entirely possible for it
to have been a problem on our end.  I would rather be pleasantly surprised that we have not made a mistake than
to leave a potentially difficult to catch bug in the code.
Collapse
Posted by Janine Ohmer on
Boy, that was quite a tangent!

Anyway, the answer to Dan's question is that I put it in the nsd startup script.  That took care of the problem.

Collapse
Posted by Jun Yamog on
Hi Dan,

Setting the Oracle profile with the LD_ASSUME_KERNEL plus putting it on nsd script should work ok.  Normally what I do is that I set the ORACLE_* and LD_ASSUME_KERNEL and run glibc21-linux-env.sh.  Putting those on init.d/oracle8i or init.d/listener8i wont be much use. Since if you have set oracle user to have the proper env it should pick that up.  Or if I have a user that I want to use any Oracle stuff like nsadmin using sqlplus I set the prover ORACLE_*, LD_ASSUME_KERNEL and glibc21-linux-env.sh.

Janine,

Its interesting that you only need to set the LD_ASSUME_KERNEL if you have the Oracle stubs patch.  I have never tried that even though I have the Oracle stubs patch.  If this is true then its a good thing.  Red Hat 8.0 does not any more have the compat libs.  I had to use the rpms from RH 7.3

In my next install I will try not to have any compat libs.  Anyway I am glad that Oracle8i and RH8 works ok for both OpenACS and CCM.  Oh yeah its seems that the latest JDBC drivers that can be downloaded in OTN does not work in oci8 mode... I had to use thin.  It seems others are experiencing this problem too.  The older drivers works fine on oci8 or thin mode.

Collapse
Posted by Dan Wickstrom on
I finally had chance to verify this, and I found that just putting the export LD_ASSUME_KERNEL statement in the nsd-oracle script was sufficient for Redhat 8.0.  Now as soon as I get the docbook tools setup (could editing the documentation be made any more difficult?), I'll modify the install docs.
Collapse
Posted by Vinod Kurup on
My latest changes to the install docs (http://66.108.4.52:8000/doc/oracle.html) do include info about using the glibc compat packages and exporting LD_ASSUME_KERNEL=2.2.5 for RH 7.3. I'll edit the nsd-oracle file in the install docs to set this parameter as well. Anyone know if this will cause Oracle to not work on other platforms?

Now as soon as I get the docbook tools setup (could editing the documentation be made any more difficult?), I'll modify the install docs.

Well, me and DocBook aren't great friends, but we tolerate each other. Do other projects have better ways of dealing with DocBook, or are you suggesting that we use a different format?

The relevant Debian packages that I have installed are:

  • docbook
  • docbook-xml
  • docbook-xsl
  • libxml2
  • libxslt1
  • psgml
  • xsltproc
The most recent time that I installed on a fresh server, I just installed those packages, edited files in emacs (psgml mode detected automatically), and generated the docs by doing 'make html' in packages/acs-core-docs/www/xml.
Collapse
Posted by Dan Wickstrom on
Well obviously, I'm not using Debian, so getting DocBook working was trial and error.  First, I searched for and installed all of the docbook rpms that I had on my install disk (actually iso images that I downloaded).  This also included installing all of the dependencies for docbook.  Next I played around tried to generate a single html file, but I kept getting an error on a undefined symbols (VERSION in this case).  Finally, I noticed the makefile and ran that, but it turned up a bunch of errors because the path names to the style sheets were wrong.  I then searched for the style sheets that were giving me errors and created symbolic links to match the paths in the error messages.  I was then able to run the makefile, and it worked, but it regenerated all of the html in acs-core-docs. Not wanting to checkin all of the these unchanged docs, I moved the acs-core-docs directory aside and  checked it out anew.  I then copied the .xml and .html files that I modified from my original directory to the newly checked out directory and checked them in.

Obviously, I've never used docbook before, but it seems that the bar for creating and updating documentation should be lower.  Something that should have taken me 30 seconds, took at least a half-hour or more.  What do we really get from using docbook other than the ablity to generate documents in other formats?  And do we really need to generate those other formats when html docs are so accessible?

Collapse
Posted by Jeff Davis on
It works pretty much out of the box on redhat 7.3. You do have to download the docbook xsl stuff but xsltproc and all the other dependencies are satisfied with a vanilla install (although there are some inconsistencies with the latest version of openacs.xsl and docbook). Maybe we should have a tarball of the docbook stuff and some clearer instructions on how to make things work. Also, I don't think anyone will fault you for making the change in the xml and just telling Roberto or Vinod about it...
Collapse
Posted by Dan Wickstrom on
For redhat 8.0, I used a vanilla install (I think they called it workstation install), and  while xsltproc was installed none of the docbook stuff was installed.  I was able to find the correct style sheets, but the rpm install puts them in a different location than expected by the current documentation build process.  A tarball with clearer instructions would probably go a long way to making it easier to update the docs.

I can't see passing all of the doc updates off to Roberto and Vinod, as that creates a bottleneck.  Not that I'm impugning their work, but I think it's better if any developer can easily update the documentation on their own.  After all, just recently, Roberto suggested we should halt all development work to fix the documentation.  Making the documentation easier to update, might help in making it so such extreme measures are not necessary.

Collapse
Posted by Jun Yamog on
Hi Vinod,

I hope that we can verify what Janine has said above.  She is able to install and run Oracle fine only LD_ASSUME_KERNEL=2.2.5 and glibc stubs patch. No more compat libs.  This is important to know since RH 8 does not anymore carry the compat libs.

In my RH 8 install I used the compat libs still since I though it was needed.  My friend at RH who installed 8i on RH 8 the minute it came out also had compat libs.

Since DanW is able to verify that we only need the LD_ASSUME_KERNEL and without running the compat libs shell script.  Which may mean that we don't need to install those compat libs.

Janine,

I am not doubting that what you have stated is wrong.  It would be nice that someone can verify it so we can remove the compat libs install on the docs.

Collapse
Posted by Janine Ohmer on
No problem, Jun - I haven't done an install on Redhat 8 and you can't assume these things will work from one release to the other (unfortunately).
Collapse
Posted by Vinod Kurup on
Well, I certainly sympathize with Dan's comments. Being defensive, I might point out that it took me (and others ) hours, if not days to get DocBook working before Roberto and I started working on the process, so if it took Dan only 30 minutes, then that is an improvement 😉

Unfortunately, I only have access to my own system, so I've tried to get things working for me, hoping that others would step in where I've screwed things up. When I start creating XSL stylesheets, I'm way at the frontiers of my realm of expertise.

I think we might be able to improve things by having the openacs.xsl sheet point to a docbook installation on the web somewhere. The index.xml sheet already does this. That way, having Docbook installed locally might not even be needed. I haven't tested this yet though.

Also, if anyone can figure out if it's possible just to generate certain desired pages instead of the whole damn thing, that would be great. I've tried without success.

I'm not going to try to defend Docbook too vigorously. It was what we were given, so I've tried to make it work. I do think it is a standard and that is a good thing. We definitely do need docs to be in HTML and PDF formats at least, judging from the multiple *ahem* "feedback" emails that I received before we had PDF docs available. I think there's also a benefit to separating content from presentation (although you need to be an XSL wizard to do it properly). But I agree with Dan that it's most important to lower the bar for documentation submissions, and I don't know the best way to do that.

Collapse
Posted by Jeff Davis on
I could set up something on openacs.org to grind the html
docs from xml from a cronjob (and make the generate log and
static html available on the web).  Does that sound like a reasonable thing to do?
Collapse
Posted by Tom Jackson on

Jeff,

That sounds like a great solution. It also took me days to get DocBook working. Once I set it down for a few months and needed to make changes to the document I created. I had already forgotten exactly what to do to get things to work. At that moment I lamented not having a simple web interface that I might have bookmarked for later use.

Collapse
Posted by Tom Jackson on

I'm going to install RH8.0 and Oracle 8.1.7 next week, assuming I can find 8.1.7. I didn't see it available anymore on the Oracle website. Where do you get copies now? (Anyone have a secret location I can get it from?)

Collapse
Posted by Dan Wickstrom on
Follow the link in the install docs to the otn site.  The link for  8i is at the bottom of the page under the heading of "Archived Products".

If we're only using docbook so that we can generate pdf docs, I have to wonder if there isn't something easier to use  I haven't researched this, but it seems that there should be tools that can be used to convert from html, or  at least xhtml, to pdf files.  That way we could maintain the original docs in html, and convert them to pdf as needed.

Collapse
Posted by Jun Yamog on
Tom,

Please try it without compat libs.  So we can get to verify if its still needed or not.  If you can't find the glibc stubs patch I can email it to you,  if you can't find it in OTN.  I also have the corrected root.sh

Also using a softlink $ORACE_HOME/ctx/bin/ctxhx -> ctxhx0 seems to be ok and removes the Intermedia problem.

Collapse
Posted by Tom Jackson on

Jun,

If you could email me the things you think I need, I would appreciate it. I'll do the install without the compat libs and see what happens. Also, I did find the 8.1.7 link on OTN.

Collapse
Posted by Jun Yamog on
Tom,

Its in your inbox.

Collapse
Posted by Tom Jackson on

Jun,

This is a preliminary update and an additional question on installing Oracle 8.1.7 on RedHat 8.0.

I followed the install doc, with the exception that I set

export LD_ASSUME_KERNEL=2.2.5
in the shell before running the runInstaller script.

That went well, and I got to the point of creating the database creation script.

When I ran that script, I must have been in a different shell, plus I had not installed the stubs patch. Since I also wasn't clear on what file that was here is what I did:

cp /mnt/cdrom/glibc-2.1.3-stubs.tar.gz.tar $ORACLE_HOME
cd $ORACLE_HOME
tar xvzf glibc-2.1.3-stubs.tar.gz.tar 
./setup_stubs.sh

This seemed to help the db creation script fail much faster! So I added the LD_ASSUME_KERNEL to my .bash_profile for the oracle user. Right now, the db creation script seems to be working.

New Question: what is the /usr/i386-glib21-linux/bin/i386-glibc21-linux-env.sh you talk about, I didn't find such a script?

Collapse
Posted by Jun Yamog on
Hi Tom,

Yup will need to to do LD_ASSUME_KERNEL everytime you make use of Oracle.  So putting it in Oracle's profile is a good thing.  Also any shell scripts that runs any Oracle stuff like nsd.

The key thing is that /usr/i386-glib21-linux/bin/i386-glibc21-linux-env.sh

should NOT be run.  We are verifying Janine's finding.  I guess you will be the one to verify that no compat libs is needed to make Oracle run on glibc2.2 based distro like RH 8.0.  So don't install any compat rpms.

Once you are successful in running Oracle with only the LD_ASSUME_KERNEL and stubs patch please post.  If you are not successful your OpenACS install should fail on CR installation.  Somewhere in SQLJ I believe.

After you verified that we can run Oracle without compat libs, we can just put a note on the Install Docs on the needed steps.

Collapse
Posted by Tom Jackson on
Once you are successful in running Oracle with only the LD_ASSUME_KERNEL and stubs patch please post. If you are not successful your OpenACS install should fail on CR installation. Somewhere in SQLJ I believe.

Woah, I ran acceptance.sql and thought I had succeeded. I'll install OpenACS today and see what happens, but assuming I fail, shouldn't the acceptance test be changed?

Collapse
Posted by Jun Yamog on
Hi Tom,

I am not really familiar what the acceptance test does.  The last time I run it was ACS 3.x days.  Since I don't know what it really provides I decided not to run anymore when I install Oracle.  I hope someone can shed good light in acceptance test.

Collapse
Posted by Tom Jackson on

Well it all crapped out at:

Loading SQLJ code /web/iunicycle/packages/acs-content-repository/java/Regexp-oracle.sqlj...

This is while trying to install the ACS Content Repository 4.6

I guess I'm back to the start, what are the compat libs, how do you install them, and where do I need to start over from?

Or, is there any way around the Content Repository? :)

The loader just stalls out with this process still running:

/ora8/m01/app/oracle/product/8.1.7/JRE/bin/linux/native_threads/jre -classpath /ora8
/m01/app/oracle/product/8.1.7/JRE/lib/rt.jar:/ora8/m01/app/oracle/product/8.1.7/JRE/lib/i18n.jar:/ora8/m01/app/
oracle/product/8.1.7/jlib/javax-ssl-1_1.jar:/ora8/m01/app/oracle/product/8.1.7/jlib/jssl-1_1.jar:/ora8/m01/app/
oracle/product/8.1.7/jdbc/lib/classes111.zip:/ora8/m01/app/oracle/product/8.1.7/sqlj/lib/translator.zip:/ora8/m
01/app/oracle/product/8.1.7/javavm/lib/aurora.zip oracle.aurora.server.tools.loadjava.LoadJavaMain -verbose -us
er iunicycle/password /web/iunicycle/packages/acs-content-repository/java/Regexp-oracle.sqlj

Note how nice this is, including the oracle password in a command line viewable via ps!

Collapse
Posted by Janine Ohmer on
Do you have LD_ASSUME_KERNEL set in your nsd startup script?  If not, adding it should fix this.
Collapse
Posted by Tom Jackson on

Yippie! It worked! :)

Thanks for all the help. Janine, apparently LS_... isn't good enough, I had a typo in my startup script. So it all works with just LD_ASSUME_KERNEL=2.2.5, and the glibc stub patch.

I really need to collect togeather the problems I ran into in locating software, adjusting scripts, etc. This installation is all but impossible, I would think, for a new user to get this right. Although the general outline is right, the few details that have changed over time make installation a real minefield. It is also so long that I can see why keeping the install document up-to-date is so difficult.

I also did something different this time around: I didn't copy software from a previous install. Believe it or not, it isn't easy locating stuff even if you know it exists.

I also recorded as many steps as possible, using a new program that records and executes commands. Oracle was the real problem, since it uses the runInstaller, but the docs are current for this section of the install.

Collapse
Posted by Jun Yamog on
Yay!

Tom please post your doc.  I want to look at it too.  I have my files, etc with me.  Maybe we can revise the docs a bit and help Vinod and Roberto out.

Collapse
Posted by Christine Stamper on
This runinstaller issue mentioned above is a documented "issue" (ahem) on Oracle's support site metalink.  Don't know what they'd think of me posting this, but here's their official word on the thing.  I ran into the same thing here, but since I have access to Oracle support through my employer, I just did a quick search.  I still had some error messages from the root.sh, but I don't think they look life threatening.  Hope this is useful:

fact: Oracle Universal Installer 8.1.7
fact: Intel Based Server LINUX
symptom: Runinstaller hangs at startup
cause: Redhat 7.1 uses newer libraries as Oracle 8.1.7 is developed with
and therefore needs a downgrade of the involved parts

fix:

To solve the initial problem of the not starting OUI you need to install the
following patches available at the Redhat 7.1 CDs or from www.redhat.com

compat-glibc-6.2-2.1.3.2.i386.rpm (CD1)
compat-egcs-6.2-1.1.2.14.i386.rpm (CD2)
compat-libs-6.2-3.i386.rpm        (CD2)

Once these patches have been installed, open new session as the oracle OS user
and type the following:
Export LD_ASSUME_KERNEL=2.2.5
Then run the following shell script:
i386-glibc21-linux-env.sh  (this can be found at /usr/i386-glibc21-linux/bin)

Now you should be able to run the OUI correctly.

When choosing your installation options do not check the OUI to create a
database.The relink fails within the installation procedure due to <bug:1467074>

When the relink fails during installation press ignore. This will take you
to the Network Configuration screen. At this point you will have to close the
OUI as it will also fail due to the relink problem.

You need to apply patch for <bug:1467074> ( as the oracle OS user)
Now you should be able to run all Oracle executables

relink utilities additionally

You will also have to put LD_ASSUME_KERNEL=2.2.5  in the profile
of the oracle OS user. So you are able to run the programs like dbassist , netca
and dbastudio

Collapse
Posted by tool who on
how exactly do i set the LD_ASSUME_KERNEL=2.2.5 in my nsd-oracle script? Thank you
Collapse
Posted by Tom Jackson on

On linux you have a profile script that you can use. In your home directory edit your .bash_profile file and add the line:

export LD_ASSUME_KERNEL=2.2.5

You could instead edit /etc/profile, using the same code.

You should put the same code in any startup script for AOLserver as well.

Note, when you do this change, you need to logout and log back in, if you are using the X desktop, you need to log out of that and log back in.

Collapse
Posted by tool who on
cool, thanks, i'll try it out.
Collapse
Posted by Mark Aufflick on
The stub patches have become quite hard to find on the oracle site. Here is a message posted by an Oracle staffer with a direct link that works.

http://www.orafaq.net/msgboard/unix/messages/457.htm