{"id":1709,"date":"2014-05-29T20:00:32","date_gmt":"2014-05-29T18:00:32","guid":{"rendered":"http:\/\/www.oradba.ch\/?p=1709"},"modified":"2014-06-10T12:41:35","modified_gmt":"2014-06-10T10:41:35","slug":"avdf-installation-fails-on-hp-server-with-smart-array-disk-controller","status":"publish","type":"post","link":"https:\/\/www.oradba.ch\/wordpress\/2014\/05\/avdf-installation-fails-on-hp-server-with-smart-array-disk-controller\/","title":{"rendered":"AVDF installation fails on HP server with Smart Array Disk Controller"},"content":{"rendered":"

I’ve successfully set up a couple of AVDF installation on different VM Server as well on HP Blade or Rack servers. On the VM server I never had any problems. For the installation of AVDF 12.1.1.x on HP servers BL465c Gen8<\/em> or DL380p Gen8<\/em>, there were always warnings during partitioning of the disks. So far it was never an issue to just continue the installation. With AVDF 12.1.2 this has changed. On some HP servers with smart array disk controller the installation fails because of problems with the drivers respectively device names.<\/p>\n

Earlier installation of AVDF like 12.1.1.3.0 simply complained about not enough space.<\/p>\n

\"AVDF_12.1.1.3.0_setup02\"<\/a><\/p>\n

OK, 0GB is a bit less for setting up an AVDF Server :-), nevertheless ignoring the error still worked. AVDF 12.1.1.2 as well AVDF 12.1.1.3 could be successfully setup using the cciss Driver for HP Smart Array. As of AVDF 12.1.2 the error is not that friendly any more.<\/p>\n

\"AVDF_12.1.2.0.0_setup02\"<\/a><\/p>\n

The title of the error “Error Parsing Kickstart Config”<\/em> indicates that there is an issue at an early stage of the system setup. It is worth having a deeper look into the kickstart configuration file. The kickstart file can be found in the initrd.img image on the AVDF installation ISO. See AVDF installation ISO<\/a> for how to extract the kickstart file.<\/p>\n

In the kickstart file we can see at line 62, that a pre-script is executed to create the partition commands. This pre-script is a python script which does create a temporary file (\/tmp\/partition-include<\/em>) with the partition commands based on the available disks. The partition command itself is then included at line 36.<\/p>\n

\r\n########## Partition the disk ##############\r\n%include \/tmp\/partition-include\r\n\r\n# Create logical volume group - this is where all volumes will reside\r\nvolgroup vg_root pv.01\r\n# Now create the volumes, a.k.a logical partitions. The data partions (\/var\/lib\/oracle) is grown\r\n# up to the specified size. The rest of the FS is left unallocated.\r\n#\r\n# You must make changes to ruby_lib\/dbfw\/dbfw_fstab.rb if you change the FS specification.\r\n#\r\nlogvol swap --fstype swap --vgname=vg_root --size=4096 --name=lv_swap\r\nlogvol \/ --fstype ext3 --fsoptions="noatime" --vgname=vg_root --size=7000 --name=lv_root\r\nlogvol \/images --fstype ext3 --fsoptions="noexec,nodev,nosuid,noatime" --vgname=vg_root --size=15000 --name=lv_images\r\nlogvol \/usr\/local\/dbfw --fstype ext3 --fsoptions="noatime" --vgname=vg_root --size=1000 --name=lv_local_dbfw\r\nlogvol \/usr\/local\/dbfw\/tmp --fstype ext3 --fsoptions="noexec,nodev,nosuid,noatime" --vgname=vg_root --size=9000 --name=lv_local_dbfw_tmp\r\nlogvol \/home --fstype ext3 --fsoptions="noexec,nodev,nosuid,noatime" --vgname=vg_root --size=1000 --name=lv_home\r\nlogvol \/tmp --fstype ext3 --fsoptions="nodev,nosuid,noatime" --vgname=vg_root --size=2000 --name=lv_tmp\r\nlogvol \/var\/log --fstype ext3 --fsoptions="noexec,nodev,nosuid,noatime" --vgname=vg_root --size=6000 --name=lv_var_log\r\nlogvol \/var\/tmp --fstype ext3 --fsoptions="noexec,nodev,nosuid,noatime" --vgname=vg_root --size=6000 --name=lv_var_tmp\r\nlogvol \/var\/www --fstype ext3 --fsoptions="nodev,nosuid,noatime" --vgname=vg_root --size=1000 --name=lv_var_www\r\nlogvol \/var\/www\/tmp --fstype ext3 --fsoptions="nodev,nosuid,noatime" --vgname=vg_root --size=1000 --name=lv_var_www_tmp\r\nlogvol \/var\/lib\/oracle --fstype ext3 --fsoptions="noatime" --vgname=vg_root --size=20000 --name=lv_oracle\r\nlogvol \/var\/dbfw --fstype ext3 --fsoptions="noatime" --vgname=vg_root --size=10000 --name=lv_var_dbfw\r\n\r\n# Tasks performed before installation\r\n%pre\r\n\r\npython \/kickstart\/partitions.py 2> \/tmp\/partitions_error\r\nif [ $? -ne 0 ]; then\r\nDISKERROR=$(\/bin\/cat \/tmp\/partitions_error)\r\nfi<\/pre>\n
Having a look into the file \/tmp\/partition-include<\/em> reveals the wrong partition command which leads to the error mentioned earlier. As you can see below the disks are specified with –ondisk=cciss!c0d0<\/em> rather than –ondisk=cciss\/c0d0<\/em>. The python script which builds the partition commands, has issues with the device names. Actually, for an HP smart array disk, the corresponding driver should be loaded so that the devices are visible as sd*. The root cause could be the missing driver or an error in the python script. I’ve opened a service request with oracle Support for further analysis.
\n\"AVDF_12.1.2.0.0_setup04\"<\/a><\/p>\n
Workarounds<\/h3>\n
For the moment I just see the following two workarounds.<\/p>\n