YaK:: Tracking down possible issues with Linux's LVM management tools [Changes]   [Calendar]   [Search]   [Index]   [PhotoTags]   

Tracking down possible issues with Linux's LVM management tools

EDIT: Syslog info added.

Text torn from my LJ entry at: http://simoncion.livejournal.com/166526.html (If anyone has run the LVM tools [specifically lvresize and pvmove] on systems with 16MB-128MB of RAM on them, and has had the tools fail, please drop me a line) [My LJ username @ gmail, or my LJ username on AIM will reach me, eventually]

I think that some of the LVM tools are flakey.

Case in point: 1) My machine has 128MB of RAM, 1GB of swap, and a Slot 1 Pentium 2 333 (running at 300). I'm running a 2.6.16 (I think) kernel, ck patchset on Gentoo Linux. I'm ssh'd into the box, doing a pvmove on a pv that's no larger than 270GB large to a pv that's ~290GB large. I let this run for quite a while (more than an hour?). The pvmove stops progressing, my system load spikes (while my CPU utilization is nil), and after another thirty minutes, I check dmesg. The kernel has oops'd with a:

"Unable to handle kernel paging request at virtual address ..."

The gkrellm daemon shows that I'm using ~70% of my RAM and very little swap. EDIT: Top told me that pvresize was using ~55% of available RAM.

The pvmove process is not responding to SIGTERM or SIGINT or SIGKILL. I can make new ssh connections, run top, and people can initiate samba transfers from the machine and all that, but when I try to kick off something like emerge, the process hangs, and stops responding to signals. (sending SIGKILL does *absolutely nothing*)

I'm in Knoppix right now, and am babysitting the pvmove so that this doesn't happen again (I hope).

2) A previous experience with lvresize, same machine, 32MB of RAM, and 1GB of swap: I would attempt to resize an lv. lvresize would silently fail segfault (IIRC). I didn't look at the dmesg output, but I'll setup a scenario in Vmware when I get settled into my new home in January.

Dear god, I'm sorry for the ramble. I'm a bit wired on caffinee at the moment. I'm babysitting the pvmove, terminating it when I'm about to run out of RAM, and restarting... : / UGH. This sucks, a lot. Why can't this program Just Work(TM)?

Anyhow, My question: Have any of you had similar experiences with the LVM tools? If so, have you resolved the issue (And how did you do it)? And, if you didn't resolve the issue, what do you think is the root cause?

If anyone wants more information... kernel config, uname output, anything really, leave a comment, and I'll try my best to provide. I'll be away from Dec 23 to Jan 3rd or so, though. NOTE: I added syslog output and other bits of information at the Livejournal post.

-Simon C. Ion

strick makes quick notes: 2011-10-29

Disk: "2010-11-N 5=pdp11" 1.5T Sata

root@retro:/z# lvcreate --size 20G --snapshot --name yakweb_snap_1 /dev/pdp11/yakweb_srv

root@retro:/z# lvcreate --size 10G --snapshot --name smilax_snap_1 /dev/pdp11/smilax_srv

(unless otherwise marked) Copyright 2002-2014 YakPeople. All rights reserved.
(last modified 2011-10-30)       [Login]
(No back references.)