[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Automatic upgrades stalls on the last kernel release



On Mon, 28 Jul 2014 15:56:52 +0100
Paul Lewis <apflewis@gmail.com> wrote:

> 
> I seem to have a problem with automatic updates on my system. The AU 
> system has been unable to function normally for some time and I now 
> have something like 61 updates waiting in the queue.
> 
> 
> When I try to process the updates, it seems to choke on the Linux 3.2 
> for 64 bit PCs entry,  which is linux-image-3.2.0-4-amd64-3.2.60-1
> +deb7u1 (64bit)
> 
> reported as 23.4MBs.
> 
> 	The error is "Failed to process request." and more details
> are -
> 
> 'cannot copy extracted data for '.lib/modules/3.2.0-4amd64/kernel/
> drivers/net/wireless/rt2x00/rt61pci.ko' to
> '/lib/modules/3.2.0-4-amd64/
> kernel/drivers/net/wireless/rt2x00/rt61pci.ko.dpkg-new'.
> 
> If I manually copy the file and rerun the update, it just produces a 
> similar error message with a different file name.
> 
> df returns the following;
> test:# df
> Filesystem                 1K-blocks    Used Available Use% Mounted on
> rootfs                        329233  270345     41890  87% /
> udev                           10240       0     10240   0% /dev
> tmpfs                         406020    1892    404128   1% /run
> /dev/mapper/test-root    329233  270345     41890  87% /
> tmpfs                           5120       4      5116   1% /run/lock
> tmpfs                         812020     148    811872   1% /run/shm
> /dev/sdb1                     233191   18789    201961   9% /boot
> /dev/mapper/test-home  39502452 5852096  31643728  16% /home
> /dev/mapper/test-tmp     376807   10303    347048   3% /tmp
> /dev/mapper/test-usr    8647944 4809964   3398684  59% /usr
> /dev/mapper/test-var    2882592  971236   1764924  36% /var
> test#
> 
> 1. I'm wondering if it's a disk space issue.
> 

Almost certainly. Kernels aren't that big, but the /lib/modules/* that
match them are now 100MB-ish. When you install a new kernel, you need
that space, but an upgrade of the same kernel release should work in
less.

> Can I purge all the update queue and try to update the kernel on its 
> own then the remaining updates.
> 

You have run out of space for something kernel-sized. Do you have an
old kernel you can get rid of? 270MB on / with separate /var and /usr
looks like two kernels-worth.

I've run into this one on my server, and will have to reorganise disc
space before the next upgrade. 300MB used to be more than enough for /,
but isn't now. I keep a spare kernel, and will have to lose it before I
upgrade to another release, though that shouldn't be a problem as I
think I would have seen any problems with the running kernel by now.

Also, a separate /usr, which has been routine until recently, is now a
bad idea, as some of it may be needed at boot time. Not yet a
show-stopper, but one day it will be. From your kernel, you are
presumably running stable, and the next stable will use systemd, with a
lot of re-jigged software.

You can nudge things around a bit without too much trouble on LVM to
get some more space on / (but it still isn't quite a trivial job,
depending on physical location and whether there is unused space or you
have to shrink something else). In the very long term, you will need to
merge / and /usr, but the trouble is that /usr must be put into /, not
the other way round, so there's a fair bit of messing about involved if
you need to minimise downtime. You can't pull /lib/modules out into
another partition, as that is very definitely required at boot time.

If you don't have an old kernel in there, you're looking at some LVM
shifting, and if your drive is four or five years old, it might be time
to consider replacement, which makes things much easier. My main
workstation drive is now five years old, and I'm waiting for its
replacement to arrive. I'll probably also replace the server drive when
I need to upgrade stable on that.

-- 
Joe


Reply to: