Wojtek Pilorz
2008-May-08 08:05 UTC
[CentOS] total server lock-up during CentOS5.1 installation with vnc
Hi, I would like to share my experience with total server lock-up during CentOS5.1 installation using vnc installation method. My setup was as follows +--------------------------------------+ | my_PC, CentOS4.6 | | X, KDE, vncviewer | | 100Mbit/s eth0 | | Intel Corporation 82801BA/BAM/CA/CAM | +------eth0----------------------------+ | +-----+-----------+ | a cheap switch | | 10/100MBit/s | +-----+-----------+ | /// goes to another floor | | | +------+------+-+----+ | 100 Mbit hub | +-----+--------------+ | | +--------------------------------------------+ | | 100Mbit/s ports (only two) | +-----+--+-------------+ | | 10/100Mbit/s switch | | +----------+-+-+-+-----+ | | | | <- 10 Mbit/s ports | | | eth0 | Intel | eth0 Intel +---------+---------------+ +--------+---------------+ | Srv1 | | Srv2 | | CentOS | crossed | Fedora6->CentOS51 | | Intel eth1 +-----------------+ eth1 Realtek | | | UTP cable | | +-------------------------+ +------------------------+ Srv1, Srv2 are almost identical Compaq ProLiant servers, each with 512MB RAM, 2*PIII CPU, Compaq Smart-2/P RAID array, two ethernet 10/100 Mbis/s controllers. Srv1: # lspci 00:00.0 Host bridge: Broadcom CNB20-LE Host Bridge (rev 04) 00:00.1 Host bridge: Broadcom CNB20-LE Host Bridge (rev 02) 00:03.0 System peripheral: Compaq Computer Corporation Advanced System Management Controller 00:05.0 VGA compatible controller: ATI Technologies Inc 3D Rage IIC 215IIC [Mach64 GT IIC] (rev 7a) 00:06.0 PCI bridge: IBM PCI to PCI Bridge (IBM27-82351) (rev 07) 00:09.0 Ethernet controller: Intel Corporation 82557/8/9 Ethernet Pro 100 (rev 05) 00:0f.0 ISA bridge: Broadcom OSB4 South Bridge (rev 4d) 00:11.0 Host bridge: Broadcom CNB20-LE Host Bridge (rev 04) 00:11.1 Host bridge: Broadcom CNB20-LE Host Bridge (rev 02) 01:00.0 Mass storage controller: Compaq Computer Corporation Smart-2/P RAID Controller (rev 03) 02:04.0 SCSI storage controller: LSI Logic / Symbios Logic 53c875 (rev 14) 02:04.1 SCSI storage controller: LSI Logic / Symbios Logic 53c875 (rev 14) 02:08.0 Ethernet controller: Intel Corporation 82557/8/9 Ethernet Pro 100 (rev 05) Srv2: # lspci 00:00.0 Host bridge: Broadcom CNB20-LE Host Bridge (rev 04) 00:00.1 Host bridge: Broadcom CNB20-LE Host Bridge (rev 02) 00:03.0 System peripheral: Compaq Computer Corporation Advanced System Management Controller 00:05.0 VGA compatible controller: ATI Technologies Inc 3D Rage IIC 215IIC [Mach64 GT IIC] (rev 7a) 00:06.0 PCI bridge: IBM PCI to PCI Bridge (IBM27-82351) (rev 07) 00:0f.0 ISA bridge: Broadcom OSB4 South Bridge (rev 4d) 00:11.0 Host bridge: Broadcom CNB20-LE Host Bridge (rev 04) 00:11.1 Host bridge: Broadcom CNB20-LE Host Bridge (rev 02) 01:00.0 Mass storage controller: Compaq Computer Corporation Smart-2/P RAID Controller (rev 03) 02:04.0 SCSI storage controller: LSI Logic / Symbios Logic 53c875 (rev 14) 02:04.1 SCSI storage controller: LSI Logic / Symbios Logic 53c875 (rev 14) 02:06.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL-8139/8139C/8139C+ (rev 10) 02:08.0 Ethernet controller: Intel Corporation 82557/8/9 Ethernet Pro 100 (rev 05) Srv1 runs CentOS5 for 8 months with no major problem. Srv2 had Fedora Core 6 installed, recently I decided to install CentOS5.1 instead. Because the servers are located in a noisy server room, have low-end CRT monitors where installation in graphical mode is not possible, I run installation using vnc method. As the servers to not have a DVD drive, I have put CentOS 5.1 DVD image on a hard disk and booted from a CentOS boot CD ( boot.iso ) or from GRUB [ using kernel and initd from boot.iso ] and specifying HardDisk installation source. First I have installed a small rescue system on /dev/ida/c0d0p1 [ Srv2 ] with eth0 as the interface vnc was using. No problem other than grub which did not install correctly and I had to install it from rescue mode using grub-install. When installing the regular system I decided to use more fancy method:>From my_PC I started ssh session to Srv1, redirecting port 5901 to Srv1/eth1/port 5901(eth1 on Srv1 and Srv2 are configured to use private IP address range) Then I specified to CentOS51 installer that vnc is going to use eth1 interface. I could connect vncviewer to localhost:1 at my_PC (that is port 5901) which was forwarded over ssh to Srv1/eth1 and Srv2/eth1. This was working, I could prepare installation in custom mode, define LVM logical volumes, mount points, etc. When I came to package selection, however, Srv2 locked hard after some time (I tries 2 or three times, always the same result, somewhere in package selection the Srv2 locked hard). By 'locked hard' I mean no response to ping over eth0 or eth1, not even ARP response on eth0 or eth1, no response to attempts to switch virtual consoles (Alt-Fx) and no message on screen which would give any clue what might have happened. After 2 or 3 attempts to install that way, I run with no problem installation using vnc over eth0 (no ssh port forwarding this time). I do not expect any help for that, just wanted to share my weird experience. (No hard data, the lockup was before any log files could be created by installer, and there was no error messages on screen). Best regards, Wojtek