web app down every three months
Joel Rees
I guess I haven't done enough homework. And I can't give enough detail
yet.
But here's the question anyway, in case the general symptoms ring a
bell with someone:
We have a web application installed at a customer's site with apache,
Oracle (8), and php on a pair of MSW2k servers. (Oracle, of course,
gets its own machine.)
About every three months (rough estimate) the web application starts
sending visitors to our "busy" page. (That's the one that catches just
about everything but attempts to jump directly to URLs we don't want
visitors jumping directly to.) The logs seem to indicate the database
is in some strange state, so the customer re-boots the database
server. (The customer does not seem to like to re-boot the server.)
I'm inclined to think this is expected behavior for the setup, myself.
If I am right, I would sure appreciate a pointer to some material to
back me up (and maybe convince my boss to move the database to a Linux
box, and apache/php to an openBSD box).
If I'm wrong, and there are some parameters we can look at to tune the
database server, I'd appreciate some suggestions.
Joel Rees
Knut Talman
> About every three months (rough estimate) the web application starts
> sending visitors to our "busy" page. (That's the one that catches just
> about everything but attempts to jump directly to URLs we don't want
> visitors jumping directly to.) The logs seem to indicate the database
> is in some strange state, so the customer re-boots the database
> server. (The customer does not seem to like to re-boot the server.)
Could you please define strange state?
Regards,
Knut
Joel Rees
Lots of time-outs, telling the program that data is not there even
though it is, that kind of thing. We haven't actually been able to
observe the symptoms ourselves, just read the logs.
Right now, I'm just thinking up bizzare ways to access the web app and
we have another guy pounding away at it. I'm also planning to write a
perl script to pound at it. I don't have a lot of experience with this
stuff, so I'm just looking for some heads-ups.
We are also checking, of course, for intrusions.
Joel Rees
Michael Austin
Is the Oracle server also on W2K? if so, Wxx has always had a history
of requiring a reboot after some time. I have been at many very large
(250+ WNT servers) that were rebooted every other week on a set schedule
for this very purpose. This "inconvenience" kept them from having
similar symptoms where the listeners started to act similar to your
situation. (And they say WNT/2K is ready for prime time -- hogwash.
If you want my professional opinion (free of charge even...) Move the
database server to a real OS. Tru64 (soon to be integrated with HP/UX),
HP/UX, or (if you have too,) Solaris. If you need it on a secure OS, use
OpenVMS.
If you need a real database engine use Oracle Rdb (not to be confused
with 7/8i/9i) on OpenVMS. I have one database that has been open
(manufacturing shop floor) for 455 days 0 hrs, 22 minutes.
I don't believe I have ever seen an Oracle database open/operational
that long. Unix, NT all need rebooting far too often to be considered
"24x7".
So, your reboot of evey 3 months is a much better track record than I
have seen from ANY site where I have worked (mostly F-500 companies).
--
Regards,
Michael Austin OpenVMS Admin/User since June 1984
First DBA Source, Inc. Registered Linux User #261163
Sr. Consultant http://www.firstdbasource.com
http://www.firstdbasource.com/donation.html
704-947-1089 (Office) 704-236-4377 (Mobile)
Joel Rees
> Joel Rees wrote:
> >
...
> > We have a web application installed at a customer's site with apache,
> > Oracle (8), and php on a pair of MSW2k servers. (Oracle, of course,
> > gets its own machine.)
> >
> > About every three months (rough estimate) the web application starts
> > sending visitors to our "busy" page. (That's the one that catches just
> > about everything but attempts to jump directly to URLs we don't want
> > visitors jumping directly to.) The logs seem to indicate the database
> > is in some strange state, so the customer re-boots the database
> > server. (The customer does not seem to like to re-boot the server.)
> >
> > I'm inclined to think this is expected behavior for the setup, myself.
> > If I am right, I would sure appreciate a pointer to some material to
> > back me up (and maybe convince my boss to move the database to a Linux
> > box, and apache/php to an openBSD box).
> >
> > If I'm wrong, and there are some parameters we can look at to tune the
> > database server, I'd appreciate some suggestions.
> >
> > Joel Rees
>
> Is the Oracle server also on W2K? if so, Wxx has always had a history
> of requiring a reboot after some time. I have been at many very large
> (250+ WNT servers) that were rebooted every other week on a set schedule
> for this very purpose. This "inconvenience" kept them from having
> similar symptoms where the listeners started to act similar to your
> situation.
This is the information I need to show my boss. This response will
help, but it would be good if I could show her some operational survey
results or something.
> (And they say WNT/2K is ready for prime time -- hogwash.
MS has been doing a pretty good job of white-washing the old
NT-needs-rebot reputation.
> If you want my professional opinion (free of charge even...) Move the
> database server to a real OS. Tru64 (soon to be integrated with HP/UX),
> HP/UX, or (if you have too,) Solaris. If you need it on a secure OS, use
> OpenVMS.
I have been wishing Oracle would port to the BSD environment as well.
Hmm. Has anyone here had much success running Oracle's Linux products
on openBSD under emulation?
> If you need a real database engine use Oracle Rdb (not to be confused
> with 7/8i/9i) on OpenVMS. I have one database that has been open
> (manufacturing shop floor) for 455 days 0 hrs, 22 minutes.
>
> I don't believe I have ever seen an Oracle database open/operational
> that long. Unix, NT all need rebooting far too often to be considered
> "24x7".
>
> So, your reboot of evey 3 months is a much better track record than I
> have seen from ANY site where I have worked (mostly F-500 companies).
Thanks for the comments. I think they'll help.
Joel Rees