Description of new feature: New feature to support swap partition as dump device. Kernel modification(FZ#156010) and diskdumputils modification are needed for swap partition support of Diskdump. Version-Release number of selected component (if applicable): RHEL4-U2
I made a patch of rc.sysinit which adds a hook to call savecore before swapon.
Tim, The code has already been included in the latest diskdumputils (1.1.7). This feature is still working correctly. Could you change the status to "MODIFIED"? Regards, Akira
This issue has already been addressed in the current RHEL4-U2 diskdumputils package that is being tested via: RHEA-2005:578 - diskdumputils enhancement update Because this BZ is has been removed from the MUSTFIX/CANFIX list for U2, the errata above is "red" flagging this bugzilla. Not only that, the errata text description contains a paragraph explaining how to use this particular feature. What should be done about this? Should all of the references to this feature be removed from the errata, considering that the functionality remains in both the diskdumputils user package and the kernel diskdump code? Please advise ASAP.
Changing state to modified as this is already in RHEL4 U2.
"/etc/init.d/diskdump swapsavecore" won't work for those who have network-mounted /usr... it isn't there during rc.sysinit, and you've got calls to /usr/sbin/diskdumpfmt and /usr/sbin/savecore in there.
Additionally, it doesn't appear likely that the use of a swap partition will work properly, due to the architecture of swap and diskdump. As it "works" now, I have to make sure that my swap partition hasn't got any data swapped into it when a crash is done, or else I'll get: CPU frozen: #0#2#3 CPU#1 is executing diskdump. start dumping <4>mptscsih: ioc0: Attempting host reset! (sc=db52b0a0) <6>mptbase: Initiating ioc0 recovery check dump partition... <3>disk_dump: bad signature in block 1 <3>disk_dump: check partition failed. <6>disk_dump: diskdump failed, fall back to trying netdump And even if there's nothing in swap and I've set up diskdump to use a swap partition (and the crash is dumped), when it comes back up, swap won't get mounted again, as the partition doesn't have the swap-space signature on it any longer. It seems that this boils down to the fact that swap requires a signature on the partition and diskdump also requires a signature (of some sort) and those are different from each other.
The swap support works on my machine. According to the above kernel message, the kernel tries to check the signature of block #1. It means that the diskdump module does not acknowledge your dump partition as a swap device. Please show the output of the following command: diskdumpfmt -vc <your dump partition> The command just checks the status of the specified device. If the dump partition is formatted as a swap partition, it will display the following: <device name> status : swap device block size : ******** note : - The following informations are helpful to solve the problem: o /etc/sysconfig/diskdump o /etc/fstab o The output of /proc/diskdump o /var/log/messages and /var/log/messages.* o The output of dmesg o The output of swapon -s Thank you,
Looks like the README needs an update, not just to include the support for a swap partition, but to make sure that you _don't_ run "/etc/init.d/diskdump initialformat" if using a swap device... as that's what I was doing (following directions) and it was failing because of it. Skipping that step makes the diskdump to swap work fine.
Jay confirmed. In U2. Removing U3Proposed.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on the solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHEA-2005-578.html