[lkml]   [2008]   [Oct]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [RFC v6][PATCH 0/9] Kernel based checkpoint/restart

* Oren Laadan <> wrote:

> These patches implement basic checkpoint-restart [CR]. This version
> (v6) supports basic tasks with simple private memory, and open files
> (regular files and directories only). Changes mainly cleanups. See
> original announcements below.

i'm wondering about the following productization aspect: it would be
very useful to applications and users if they knew whether it is safe to
checkpoint a given app. I.e. whether that app has any state that cannot
be stored/restored yet.

Once we can do that, if the kernel can reliably tell whether it can
safely checkpoint an application, we could start adding a kernel driven
self-test of sorts: a self-propelled kernel feature that would
transparently try to checkpoint various applications as it goes, and
restore them immediately.

When such a test-kernel is booted then all that should be visible is an
occasional slowdown due to the random save/restore cycles of various
processes - but no actual application breakage should ever occur, and
the kernel must not crash either. This would work a bit like
CONFIG_RCUTORTURE: a constant test that should be transparent in terms
of functionality.

Also, the ability to tell whether a process can be safely checkpointed
would allow apps to rely on it - they cannot accidentally use some
kernel feature that is not saved/restored and then lose state across a
CR cycle.

Plus, as a bonus, the inability to CR a given application would sure
spur the development of proper checkpointing of that given kernel state.
We could print some once-per-boot debug warning about exactly what bit
cannot be checkpointed yet. This would create proper pressure from
actual users of CR.


 \ /
  Last update: 2008-10-09 14:51    [W:0.186 / U:0.164 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site