missing c++ lib when installing

[n8behavior]

the install script on my RHEL 4 says I am missing compat-libstdc++-296. before installing i did

yum install compat-libstdc++-33
yum install compat-libstdc++-296

both completed without error. ls -l |grep libstdc++ yield this

-rwxr-xr-x   1 root root  258288 Feb 21  2005 libstdc++-2-libc6.1-1-2.9.0.so
-rwxr-xr-x   1 root root  268460 Feb 21  2005 libstdc++-3-libc6.2-2-2.10.0.so
lrwxrwxrwx   1 root root      30 Aug 21 23:40 libstdc++-libc6.1-1.so.2 -> libstdc++-2-libc6.1-1-2.9.0.so
lrwxrwxrwx   1 root root      31 Aug 21 23:40 libstdc++-libc6.2-2.so.3 -> libstdc++-3-libc6.2-2-2.10.0.so
lrwxrwxrwx   1 root root      18 Aug 21 23:34 libstdc++.so.5 -> libstdc++.so.5.0.7
-rwxr-xr-x   1 root root  733488 Feb 21  2005 libstdc++.so.5.0.7
lrwxrwxrwx   1 root root      18 Aug 21 22:46 libstdc++.so.6 -> libstdc++.so.6.0.3
-rwxr-xr-x   1 root root  792856 May  2 09:58 libstdc++.so.6.0.3

so, am i really missing the libs or is something else going on?

[mmorse]

Welcome to the forums,
This wouldn't happen to be RHEL 4 64bit would it?

If so let's check what architectures you have installed:

 rpm -qa --queryformat "%{name}.%{version}.%{arch}\n" |grep compat

[Klug]

If you're running RHEL 64 bits as mmorse suggested, you need to add the 32 bits library :

up2date compat-libstdc++-33 --arch=i386

[n8behavior]

hmmm...not at work yet, so can't get to the box but...it is a 32bit kernel- it's running on an old 2.8GHz intel celeron...my test box. however, it is a somewhat weird setup.

this is a RHEL 5 (32bit) system running the latest OpenVZ kernel. i am trying to get zimbra running in a virtual environment (VE). the VE is based on a RHEL4 template. so this could be the problem.

honestly, i was hoping it would be a common, silly mistake that everyone makes i will poke at it some more once i get to work and post back what i find. BTW, once i get Zimbra working with OpenVZ, i plan to write up a HowTo. is there a place for those on the Zimbra site?

[mmorse]

yup, the wiki
Not to detract from another open-source effort (OpenVZ) but you might also play with xen, which is also open-sourse but it's already bundled with RHEL.

[n8behavior]

yea, Zen is great! we use that and we have a VMWare ESX production environment. my attraction to OpenVZ is that the VE (really just a container) is much lighter than VMs based on VMWare or Zen from a resource POV. VMWare really eats up the resources quickly. Even with Zen each VM's kernel requires its own memory space. i am curious how Zimbra will do in a light weight container.

i'll keep posting here with updates. and i will put my final write up the wiki as you mentioned. thanks!

[mmorse]

Ya, sometimes I'm forced into 8 or 10 GB/box...any update on your missing compat-libstdc++-296 issue?

[n8behavior]

okay, i am past the "missing lib" issue. it is something i do not understand fully and will investigate later when polishing the install procedure.

under OpenVZ you usually install packages into the VEs using the command 'vzyum.' this is great for bulk admin e.g. installing a package into 400 VEs is a simple loop in a script. however, for some reason...doing this

vzyum install compat-libstdc++-296
vzyum install compat-libstdc++-33

on the host node, installs the second package fine, but not the first...even though yum says it IS installed. after entering into the VE and doing a normal yum install, the lib is now installed and the ZCS install continues past this stage.

now i am fixing up the VE counters (e.g. mem, CPU & disk quotas) to allow the installer to finish without error.

more to come...

[n8behavior]

the problem now is installation of zimbra-core RPM fails. perhaps, the problem is that the kernel i am running on the host node (HN) is not compatible with the RPM.

$ uname -r
2.6.18-8.1.8.el5.028stab039.1

the NH is CentOS 5 and runs the OpenVZ kernel. the virtual environment (VE), which shares the same kernel, is based on a CentOS 4 template i.e. it uses the CentOS 4 yum repos. so i tried ZCS for RHEL 4. any guess why the RPMs fail to install. i can install every other RHEL4/CentOS4 RPM i have tried.

i could rebuild the NH to CentOS 4, but would like the newer kernel in CentOS 5 for other VEs. or i could make the VE for ZCS based on CentOS 5 even though these templates are quite new and officially offered by OpenVZ--they are contributed and some what hacked

having to share a kernel for all VEs is downside of OS virtualization compared to para- or full-virtualization--it is would you trade for low resource overhead and speed.