Peter Selby
2013-Mar-22 00:24 UTC
Experiencing kernel panic with SSH + LVM + Xen + significant amounts of data
Hi all, I''m running into a kernel panic with Xen, under fairly specific circumstances. I''ll post it here first, and follow up on xen-devel if it looks like it''s a Xen bug. After some experimentation, it seems like the panic occurs under the following circumstances: 1) A significant amount of data is being moved 2) ...Onto (or off of?) an LVM partition 3) ...With SSH being involved somehow? 4) ...On Xen. Originally, I was trying to move a filesystem over the network via "dd if=... | ssh dd of=...", onto an LVM partition. The filesystem was 20G, and it repeatably crashed, usually after >25% transferred. I''ve also tried scp''ing files, using dd to transfer files, transferring the 20G FS to /dev/null, moving data OFF the host in question, etc. None of these triggered a panic. Moving large files on the host itself doesn''t trigger a crash. However, piping data to an LVM partition (and in one case, piping from LVM to /dev/null) has triggered a crash, though the panic seems to occur simultaneously with an SSH login when it''s happened. Minus an SSH login, I''ve successfully and repeatably moved data from LVM => LVM, LVM => null, yes => LVM, without a crash (logged in via SSH the whole time, I should add...there seems to be something about logins specifically) The system has ended up in a bad state more than once, too, which complicates things. It''s possible that the crash triggered by moving OFF of LVM to /dev/null was the result of a bad state triggered earlier, because I wasn''t able to reproduce it. No such issues occur if I''m not running Xen, which seems to rule out HW failure. To be clear, the host in question is the hypervisor, and the panics occur regardless of whether VMs are running. Version info: Essentially, it''s a fully up-to-date Ubuntu 12.04 system. HW - HP ProLiant DL145 G2/K85NL (rackmount) Xen - 4.1.2-2ubuntu2.6 LVM: LVM version: 2.02.66(2) (2010-05-20) Library version: 1.02.48 (2010-05-20) Driver version: 4.22.0 SSH: OpenSSH_5.9p1 Debian-5ubuntu1, OpenSSL 1.0.1 14 Mar 2012 Kernel - 3.2.0-39-generic I can provide more info on request. I''ve also attached a sample kernel panic. This one''s a little old, but the crashes don''t always dump to syslog, and the symptoms haven''t changed. Thanks! Peter _______________________________________________ Xen-users mailing list Xen-users@lists.xen.org http://lists.xen.org/xen-users
Aaron Cossey
2013-Mar-22 08:14 UTC
Re: Experiencing kernel panic with SSH + LVM + Xen + significant amounts of data
I can''t help with the Kernel Panic but I can say that I too used to copy LV''s around with "dd if=... | ssh dd of=...", a much more elegeant and fault tolerant way to do these copies is to export an LV as an iSCSI Target, attach the Target on the remote machine and then dd locally. Aaron Cossey aaron.cossey@gmail.com On Fri, Mar 22, 2013 at 1:24 AM, Peter Selby <peter@selby.ca> wrote:> Hi all, > > I''m running into a kernel panic with Xen, under fairly specific > circumstances. I''ll post it here first, and follow up on xen-devel if > it looks like it''s a Xen bug. > > After some experimentation, it seems like the panic occurs under the > following circumstances: > > 1) A significant amount of data is being moved > > 2) ...Onto (or off of?) an LVM partition > > 3) ...With SSH being involved somehow? > > 4) ...On Xen. > > Originally, I was trying to move a filesystem over the network via "dd > if=... | ssh dd of=...", onto an LVM partition. The filesystem was > 20G, and it repeatably crashed, usually after >25% transferred. > > I''ve also tried scp''ing files, using dd to transfer files, > transferring the 20G FS to /dev/null, moving data OFF the host in > question, etc. None of these triggered a panic. > > Moving large files on the host itself doesn''t trigger a > crash. However, piping data to an LVM partition (and in one case, > piping from LVM to /dev/null) has triggered a crash, though the panic > seems to occur simultaneously with an SSH login when it''s > happened. Minus an SSH login, I''ve successfully and repeatably moved > data from LVM => LVM, LVM => null, yes => LVM, without a crash (logged > in via SSH the whole time, I should add...there seems to be something > about logins specifically) > > The system has ended up in a bad state more than once, too, which > complicates things. It''s possible that the crash triggered by moving > OFF of LVM to /dev/null was the result of a bad state triggered > earlier, because I wasn''t able to reproduce it. > > No such issues occur if I''m not running Xen, which seems to rule out > HW failure. To be clear, the host in question is the hypervisor, and > the panics occur regardless of whether VMs are running. > > Version info: > > Essentially, it''s a fully up-to-date Ubuntu 12.04 system. > > HW - HP ProLiant DL145 G2/K85NL (rackmount) > > Xen - 4.1.2-2ubuntu2.6 > > LVM: > LVM version: 2.02.66(2) (2010-05-20) > Library version: 1.02.48 (2010-05-20) > Driver version: 4.22.0 > > SSH: > OpenSSH_5.9p1 Debian-5ubuntu1, OpenSSL 1.0.1 14 Mar 2012 > > Kernel - 3.2.0-39-generic > > I can provide more info on request. > > I''ve also attached a sample kernel panic. This one''s a little old, but > the crashes don''t always dump to syslog, and the symptoms haven''t > changed. > > Thanks! > > Peter > _______________________________________________ > Xen-users mailing list > Xen-users@lists.xen.org > http://lists.xen.org/xen-users >_______________________________________________ Xen-users mailing list Xen-users@lists.xen.org http://lists.xen.org/xen-users
Reasonably Related Threads
- Task blocked for more than 120 seconds.
- wct4xxp Excessive Interrupts Resulting in Unusable System or Card
- [LLVMdev] Heads up, I've backed out significant amounts of the multiple address space conversion changes
- [LLVMdev] Heads up, I've backed out significant amounts of the multiple address space conversion changes
- [LLVMdev] Heads up, I've backed out significant amounts of the multiple address space conversion changes