Forum OpenACS Q&A: Anyone have trouble with nsd on RedHat 7.2?

Slightly off-topic, but it is ACS-related. ;)

We (furfly) just took over hosting of Archnet, which is one of the
very early sites aD built for MIT.  The site still requires nsd 2.3.3,
because it uses a number of the /NS routines.

We switched the site from an older Solaris box at MIT to our brand
spanking new RedHat 7.2 server yesterday, and it tripped and fell flat
on it's face.  Near as I can tell, threads are not being released
properly so it just keeps creating more and more until it has reached
it's limit and they are all "in use".  Then everyone gets either a
"server busy" response or none at all, even though the load on the
machine is 0.0 and nothing at all is happening.

We spoke to the folks who sys admin, thinking that it might
also be using an older nsd and perhaps they had found a workaround.
Turns out they're using 3.3+ad13 but they also had some problems on
RedHat 7.2, though different (it would just *stop*, and zombie
processes would interfere with restarting the server and even
rebooting the system).

At this point the only solution I have (short of moving the site back
to Solaris) is working with the site's programmer to upgrade it to
nsd3.3, but I'm a bit concerned that it still may not work.  Of
course, we're talking about a site which is way, way, way less busy
than, and the hardware is a bit old and crufty, so
it may be that their experience just does not apply.

Hence the reason for this post - can anyone reassure me that they are
running 3.3+ad13 on a moderately busy site on RedHat 7.2 without major


Not a direct answer to your question but we are using AOLserver/3.3.1+ad13 on RedHat 6.0 for over a year without trouble on a moderately busy site.  Occasionally we restart aolserver but only
'cause of changes to /tcl.  I heard that RH 6.0 wasn't the best
but it's hard to knock success.


Posted by Janine Ohmer on
Thanks, Bob!

I should mention that we have plenty of sites which run just fine on RedHat 6.2 (plus about a gazillion security patches), so the problem, if there is one, is something specific to 7.2.  Or 7.x, anyway.

Don't know if it helps but we've got a number of installations on Mandrake 7.2, 8 and 8.1 and 8.2 and as this distro shares a common kernel, and is really, really close to RH7.2, that may indicate it should be ok.

However, in general Mandrake has been much better than RedHat. Given that Mandrake is 100% RH compatible, perhaps this distro might be a better alternative. 8.2 is our current distro of choice.


Hello Janine,

I am not running a production site yet, but I am running two development servers, one on RedHat 7.1 and the other one on 7.2 and both never presented the symptoms mentionned. They don't eat up memory, well, not more than usual ;-).

I did run some stress tests (with ab and an evaluation version of e-load) and everything went fine.

Hope this helps.

Just to clarify, the mention of eating up memory is not related to the symptoms you mention, just as a side-note, just to mention that I don't see (at least now) memory leaks. Using 3.3+ad13
Posted by David Walker on
We've been running on Redhat 7.0 for over a year with no difficulties.  I'm
not inclined to upgrade without cause.

I've installed Redhat 7.2 on one non-production box so I can't speak for it.

Posted by mark dalrymple on
2.3.3 is *very* sensitive to the glibc version you have.  That was our biggest Linux support headache.  Things would work groovily on one version, then break in weird and wonderful ways in the next (mainly since nsd stressed the thread safety of the library)
I have used in one of our servers, wich has a:
Celeron 600Mhz
Running RedHat 7.1
Oracle 8i
aolserver 3.2 & 3.3 + ad stuff
and a couple of middle sized sites in production state.

I haven't had problems like you said, I would say that we have been running with this configuration for more than one year and I won't change it, anyway, why you don't try quickly with redhat 7.1, maybe is a specific OS problem? I really don't know.
Posted by Janine Ohmer on
To follow up, the developer and I did a hasty upgrade to 3.3+ad13 and the site's running fine now.  I think it's pretty safe to say that 2.3.3 is truly at end-of-life now.
Posted by Cynthia Kiser on
I have a site that is kind of in the same boat as Janine's - older site (ACS 2.1 based) running on AOLServer 2.3.3 with some dependencies on NS functionality. It is on a Solaris 5.7 machine and was running happily against Oracle 8.1.5. When the Oracle instance was upgraded to 8.1.7, we started having problems with connections accumulating - accumulating to the point that we run out of them and we needed to restart Oracle to clear them out. I was going to try just upgrading the Oracle driver - we are currently running 1.0.3 - but the 2.5 driver uses TclFree - a Tcl 8 function. Anyone have guesses as to what is going on?

Migrating to a newer AOLServer is possible but would mean giving up some edit pages that use the NS/Db functions of 2.3.3 - not to mention finding all of the regexps on the site. (Janine, any other things I should put on my migration fix list if it comes to that?)

Most (maybe all) of the /NS urls are implemented in tcl ($NS_HOME/modules/tcl), the /NS/Db ones are in $NS_HOME/modules/tcl/nsdb).  Depenending on your pain threshold, you could port those to aolserver 3 land. (and maybe contribute them back to the aolserver folks, or give 'em to the open acs folks) since there are some useful utilites in there (and some big security holes if you're not careful)