Preserving uptime with kexec?

Preserving uptime with kexec?

[Thomas Tuttle]

[-- Attachment #1: Type: text/plain, Size: 635 bytes --]

Like many people, I like to brag about how great my uptime is.  But like
many other people, I like to keep my kernel up-to-date with the latest
and greatest from kernel.org.  I recently discovered the magic of kexec,
which allows me to switch kernels without rebooting for real.
Unfortunately, kexec resets my uptime when it runs.

Would anyone be interested in fixing this, and/or would anyone be
interested in *me* writing a patch for it?  I don't know if it violates
some contract where uptime is counted from the kernel boot, but it seems
like, for consistency's sake, it should count from the last hardware
boot.

--Thomas Tuttle

[-- Attachment #2: Type: application/pgp-signature, Size: 191 bytes --]

Re: Preserving uptime with kexec?

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 799 bytes --]

On Mon, 31 Jul 2006 08:59:13 EDT, Thomas Tuttle said:
> Like many people, I like to brag about how great my uptime is.  But like
> many other people, I like to keep my kernel up-to-date with the latest
> and greatest from kernel.org.  I recently discovered the magic of kexec,
> which allows me to switch kernels without rebooting for real.
> Unfortunately, kexec resets my uptime when it runs.

The reset of uptime is probably a Good Thing.  Consider the case of
a kernel memory leak - you look in /proc/meminfo and find that you've managed
to lose 64 meg of memory to the leak.  Where you start looking for the
leak will depend on whether it's 64 meg lost across 4 weeks since the
last boot, or the 30 minutes since the last boot.

(Speaking as somebody who's run into both classes of leaks...)



[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: Preserving uptime with kexec?

[Nick Warne]

Speaking of which, I have submitted a claim to Guiness Book of Records
for my uptime on a lowly 486 box that serves my webpages (via NFS) as
a 'home user' (I am sure business classed machines do better with UPS
etc.).  I have posted here twice when it hit 1000 days and then 1500
days:

[nick@486Linux nick]$ uname -a
Linux 486Linux 2.2.13-7mdk #1 Wed Sep 15 18:02:18 CEST 1999 i486 unknown

[nick@486Linux nick]$ last -xf /var/run/utmp runlevel
runlevel (to lvl 3)                    Sun Oct 14 16:07 - 16:42 (1751+00:34)

The claim has been accepted, and is now in evaluation... so I dunno
what happens next until I hear from them (or how it can be verified
unless someone pops along and I telnet into the box for them).

Nick



On 31/07/06, Valdis.Kletnieks@vt.edu <Valdis.Kletnieks@vt.edu> wrote:
> On Mon, 31 Jul 2006 08:59:13 EDT, Thomas Tuttle said:
> > Like many people, I like to brag about how great my uptime is.  But like
> > many other people, I like to keep my kernel up-to-date with the latest
> > and greatest from kernel.org.  I recently discovered the magic of kexec,
> > which allows me to switch kernels without rebooting for real.
> > Unfortunately, kexec resets my uptime when it runs.
>
> The reset of uptime is probably a Good Thing.  Consider the case of
> a kernel memory leak - you look in /proc/meminfo and find that you've
> managed
> to lose 64 meg of memory to the leak.  Where you start looking for the
> leak will depend on whether it's 64 meg lost across 4 weeks since the
> last boot, or the 30 minutes since the last boot.
>
> (Speaking as somebody who's run into both classes of leaks...)
>
>
>
>

Re: Preserving uptime with kexec?

[Jan Engelhardt]

>
> Speaking of which, I have submitted a claim to Guiness Book of Records
> for my uptime on a lowly 486 box that serves my webpages (via NFS) as
> a 'home user' (I am sure business classed machines do better with UPS
> etc.).  I have posted here twice when it hit 1000 days and then 1500
> days:

http://en.uptime-project.net/
someone's going to beat you by lengths...


Jan Engelhardt
-- 

Re: Preserving uptime with kexec?

[Nick Warne]

On Monday 31 July 2006 17:28, Jan Engelhardt wrote:
> > Speaking of which, I have submitted a claim to Guiness Book of Records
> > for my uptime on a lowly 486 box that serves my webpages (via NFS) as
> > a 'home user' (I am sure business classed machines do better with UPS
> > etc.).  I have posted here twice when it hit 1000 days and then 1500
> > days:
>
> http://en.uptime-project.net/
> someone's going to beat you by lengths...

Yeah... but how do you prove it?  The top machine on http://counter.li.org/

http://counter.li.org/reports/uptimestats.php

is just above mine - but running a kernel that wasn't even released that 
amount of days ago?

Anyway, on cue (would you believe THIS?!?), about 10 minutes after I replied 
to this thread, I received a mail:




Claim ID: xxxxxx
Membership Number: xxxxx

Dear Mr Warne,

Thank you for sending us the details of your recent record proposal for 'home 
computer - longest 'uptime''. We are afraid to say that we are unable to 
accept this as a Guinness World Record.

We have considered your proposal carefully but regret that it is not something 
for which we are currently interested in listing a record. We receive over 
60,000 enquiries a year from which only a small proportion are approved by 
our experienced researchers to establish new categories.

We realize that this will be disappointing to you. However, we have considered 
your proposal fully; in the context of the specific subject area and that of 
records as a whole, and our decision is final in this matter.

Once again thank you for your interest in Guinness World Records.

Yours sincerely,

Amarilis Espinoza
Records Management Team




Heh.  BUGGER!

Nick
-- 
Every program has two purposes:
one for which it was written and another for which it wasn't.