Topic: Issue with SME updates

[Lucky Dragon]

I've recently been trying to upgrade my dad's old file/print server (running SME 7.5.1). The machine it was on is an old P4 with 256MB RAM and a 30GB HDD, which had only 2GB free. He recently upgraded his workstation and so I thought to use his old PC to replace the old SME box and install v8 while I was at it.
Motherboard: ASUS A7N8X-X (No, the workstation isn't exactly cutting edge either)
Processor: Athlon XP @ 1.2 GHz
Memory: 1 GB
Since the motherboard doesn't have an onboard SATA controller, I put in a SiI 3114 card I had handy and attached 2 500GB drives to it (for mirroring)
I'm using the onboard nForce NIC for the WAN connection and an Intel Pro/1000 GT PCI adapter for the LAN side.

The install went smoothly off the CD, I then configured it (gateway/private server) for his setup and it seemed to be working fine. A week later I happened to check the software updates on the web interface and noticed there were some updates available to download, so I went ahead and did that... and this is when the problems started to crop up. He called me the next day and said his laptop couldn't connect to the network. When I checked the system message log, it showed that the laptop wasn't being assigned an IP address, and while I was working on that, his workstation (wired connection) started having the same problem. I hooked his old server back up and took this one back to my place for further testing. I also noticed startup error messages from SQL at this point. Since all his data was backed up, I figured the easiest thing to do would be to just wipe everything and reinstall fresh. I did this 4 times, and each time, everything went fine until I tried to download the updates, after which the system would become unstable and/or completely non-functional (kernel panic on boot). Given the errors I was seeing, I tried installing all the updates except the following:

bind-libs.i386 30:9.3.6-20.P1.el5_8.1 - smeupdates
bind-utils.i386 30:9.3.6-20.P1.el5_8.1 - smeupdates
postgresql-libs.i386 8.1.23-5.el5_8 - updates

Since then, I have had the system hooked up and running at my place for several days with no issues. I searched the forums and bugtracker for A7N8X and Silicon Image 3114 but don't see anything about compatibility issues in either place. Should I open a bug report or is there something else here I am missing?

[Jáder]

I'd open a bug BUT PostgreSQL is not standard on SME...not sure about that postgresql-lib package...

[janet]

Lucky Dragon

I think you need to tell us what repos are enabled, what contribs are installed and what configuration changes you have made after doing the install from CD. If you have lodged a bug report please advise what bug number, and keep all follow up information in the one place ie in the bug report, thanks.

[Lucky Dragon]

No contribs are installed and no unusual configuration changes: just standard items like adding a couple of user IDs and an ibay, changing the Samba workgroup name; all changes made through the web configuration interface. As for updates listed, I'll have to do a fresh install and get back to you with the complete list of updates that are displayed on the web management interface at that stage. I guess I can install those one at time until it fails if that would help narrow things down more.

This is the 2nd install of SME 8.0 I've done, and the other machine (with a somewhat newer motherboard, but also has an nForce chipset) is working fine, which is what led me to suspect a hardware incompatibility.

Edit: After a fresh install off the CD, here is the complete list of packages displayed when I go to Configuration > Software installer on the SME web management console. The default settings here are Check for updates: Enabled, Manage individual packages: Disabled. I click the button labeled Install available updates and get the following list of packages:

Code: [Select]

bind-libs.i386 30:9.3.6-20 P1.el5_8.1 - smeupdates
bind-utils.i386 30:9.3.6-20 P1.el5_8.1 - smeupdates
e2fsprogs-libs.i386 1.39-34.el5_8.1 - updates
expat.i386 1.95.8-11.el5_8 - updates
httpd i386 2.2.3-65.el5.centos - smeupdates
kernel.i686 2.6.18-308.11.1.el5 - updates
libtiff.i386 3.8.2-15.el5_8 - updates
logrotate.i386 3.7.4-14 - smeupdates
man.i386 1.6d-3.el5 - smeupdates
mod_ssl.i386 1:2.2.3-65.el5.centos - smeupdates
openssl.i686 0.9.8e-22.el5_8.4 - smeupdates
postgresql-libs.i386 8.1.23-5.el5.el5_8 - updates
python-libs.i386 8.1.23-5.el5_8 - updates
python.i386 2.4.3-46.el5_8.2 - updates
tzdata.i386 2012c-1.el5 - smeupdates
This list appears to be the same as what I would get by entering  yum list updates from the command line. (incidentally, I get "error: rpmdb: damaged header #83 retrieved -- skipping." just before it outputs the above list... should I be concerned about that?) Given jader's comment above, are there any packages displayed there that are non-standard and should be skipped? So far, I have installed only the kernel.i686 2.6.18-308.11.1.el5 - updates package. That, at least seems to not be causing any issues.
On a related note, I found this page, and using the command rpm -qa --last | more gets me a date/time sorted list. I was thinking maybe I could use yum remove <packagename> to back out any of these updates after I trip over the one(s) causing my issue. Is that feasible or am I off-base?

[Lucky Dragon]

Given the inconsistent nature of the errors and problems I was encountering after running the update from the web console, I decided to remove the SATA controller from the equation and see if the problems went away. I pulled the controller card and hooked up a PATA drive to the motherboard's integrated IDE controller, then wiped it and installed and configured a fresh copy of SME 8.0. The install went smoothly and completed with no issues. Then I went into the web console and ran the update with everything selected. The following packages were downloaded and installed:

Code: [Select]

Installed:
  kernel.i686 0:2.6.18-308.11.1.el5                                             

Updated:
  bind-libs.i386 30:9.3.6-20.P1.el5_8.1  bind-utils.i386 30:9.3.6-20.P1.el5_8.1
  e2fsprogs.i386 0:1.39-34.el5_8.1       e2fsprogs-libs.i386 0:1.39-34.el5_8.1 
  expat.i386 0:1.95.8-11.el5_8           httpd.i386 0:2.2.3-65.el5.centos       
  libtiff.i386 0:3.8.2-15.el5_8          logrotate.i386 0:3.7.4-14             
  man.i386 0:1.6d-3.el5                  mod_ssl.i386 1:2.2.3-65.el5.centos     
  openssl.i686 0:0.9.8e-22.el5_8.4       postgresql-libs.i386 0:8.1.23-5.el5_8 
  python.i386 0:2.4.3-46.el5_8.2         python-libs.i386 0:2.4.3-46.el5_8.2   
  tzdata.i386 0:2012c-1.el5
No unusual errors or messages appeared in the output during this process, and after the reboot and reconfigure the system came up clean and has been running without any errors since early this afternoon. To me this would seem to indicate some sort of conflict between the controller card and one of the updated packages. Prior to installing SME on this system, it was running XP off of a drive attached to that controller, and that was working OK, so I don't think the card itself is defective. I searched the bug tracker for issues with the 3114 controller but didn't find anything--in fact I found a couple of mentions of people using 3112 and 3114 controller cards with earlier versions of SME on the forums--so it could be that the particular card I have is a little bit off the standard spec for the Linux driver in some way.
Thanks for the response on this. If you folks think it'd be useful I will open a bug report with the current information, but I picked up a newer used motherboard from a local computer repair shop with onboard SATA support for under $30 and I should be able to get that working without any more hassle.