Red Hat Bugzilla – Bug 476001
[RHEL5.2] kdump to iSCSI device fails
Last modified: 2009-02-17 13:00:54 EST
kdump to iSCSI device fails because the driver for iSCSI does not exist in initrd.img. The initscripts in the initrd.img generated for kdump also lacks the functions to + load the device driver for iSCSI + starting up the network, and + connecting to the iSCSI device. Version-Release number of selected component (if applicable): RHEL5.2 GA kexec-tools-1.102pre-21.el5 How reproducible: Always Steps to Reproduce: 1. Install RHEL5.2 GA and reboot. 2. Add the setting, which dump is saved to iSCSI device, to "/etc/kdump.conf". 3. Add "crashkernel=128M@16M" to kernel parameter. 4. Enable the kdump service by the following command. # chkconfig kdump on 5. Reboot the system. Note: Make kdump service effective manually after the system boots up when booted from iSCSI device. 6. Make a panic by the following command. # echo c > /proc/sysrq-trigger Actual results: kdump to iSCSI device fails. See attached serial console log(rhel52x86-kdump-console.log) for detals. Expected results: kdump to iSCSI device succeeds. Hardware info: Server : Express5800/120Bb-6 OS : RHEL5.2 GA Kernel : 2.6.18-92.el5PAE NIC : Intel 82571EB Gigabit Ethernet Controller iSCSI : iStorage E1-10 HDD : SATA 750GB x 4 The above-mentioned server is booted from the attached iSCSI device. See attached sosreport for details. Additional info: - /etc/mkinitrd includes proper support for iSCSI devices. /sbin/mkdumprd should also be changed accordingly. - Provided a patch based on mkinitrd to include iSCSI support in /sbin/mkdumprd
Created attachment 326623 [details] suggested patch to add iSCSI support
Created attachment 326625 [details] console log with original mkdumprd failing with iSCSI devices
This is hardly a high priority bug. We've lived without iscsi support since GA, we don't need to cram it in at the very last minute for 5.3, especially not such a large and complex patch As far as the patch goes, It looks ok, although it looks like you've duplicated some function. As for the iscsi bits, it looks ok. I'll look more closely at this for 5.4
I just reviewed this patch, and its just wrong. It hacks in support for iscsi without any regard to how the other targets are handled. I think this is going to have to be devel nak from me. I've got an RFE for 5.4 for various other support requests, for which I'm developing an advanced configuration mode. You'll be able to handle iscsi support from there. This particular request is a WONTFIX from me