Re: lease.openlogging.org is unreachable

Re: lease.openlogging.org is unreachable

[James Bottomley]

> The interesting thing is that the code already has a backup in it and I just
> checked that code path and it works.
> 
> Has anyone else been shut down because of lease.openlogging.org being down
> and if so what version of BK were you running please?

I believe I've reported this problem to you and support@bitmover.com
several times.

There's something in BK that refuses to work when it can't contact
lease.openlogging.org, regardless of whether you just renewed the lease
or not.  This keeps biting me when I try to use BK disconnected from the
internet (usually while travelling).

This problem is certainly my single biggest headache with BK since it
means I can't go through my backlog of email and build the SCSI BK tree
when I'm travelling (which is often the only time I have available for
doing it).

My current version is bk 3.2.3 but it's been a problem on all prior
versions I've used as well.

James

Re: lease.openlogging.org is unreachable

[Larry McVoy]

On Sun, Dec 26, 2004 at 10:12:10AM -0600, James Bottomley wrote:
> > The interesting thing is that the code already has a backup in it and I just
> > checked that code path and it works.
> > 
> > Has anyone else been shut down because of lease.openlogging.org being down
> > and if so what version of BK were you running please?
> 
> I believe I've reported this problem to you and support@bitmover.com
> several times.
> 
> There's something in BK that refuses to work when it can't contact
> lease.openlogging.org, regardless of whether you just renewed the lease
> or not.  This keeps biting me when I try to use BK disconnected from the
> internet (usually while travelling).

I suspect that your hostname changes when you disconnect.  Leases are 
issued on a per host basis.  If you make your hostname constant when
you unplug it should work.  If it doesn't, let us know.
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: lease.openlogging.org is unreachable

[James Bottomley]

On Sun, 2004-12-26 at 08:27 -0800, Larry McVoy wrote:
> I suspect that your hostname changes when you disconnect.  Leases are 
> issued on a per host basis.  If you make your hostname constant when
> you unplug it should work.  If it doesn't, let us know.

Well, that's a new one, but no, I have a fixed hostname which dhcp is
forbidden from changing.

James

[BK] disconnected operation

[Larry McVoy]

On Sun, Dec 26, 2004 at 10:43:13AM -0600, James Bottomley wrote:
> On Sun, 2004-12-26 at 08:27 -0800, Larry McVoy wrote:
> > I suspect that your hostname changes when you disconnect.  Leases are 
> > issued on a per host basis.  If you make your hostname constant when
> > you unplug it should work.  If it doesn't, let us know.
> 
> Well, that's a new one, but no, I have a fixed hostname which dhcp is
> forbidden from changing.

Let's do a little poll here to find out if it is specific to you or if
this is a problem that everyone is having.  Could we get people who
use BK disconnected to stand up and be counted?  Does this work for 
anyone?

For James, could you do a little debugging please?  Run the following
when you are plugged in and it works and also when it doesn't:

	bk getuser
	bk getuser -r
	bk gethost
	bk gethost -r
	bk dotbk

We'll track it down and fix it if it is a problem on our end.  This stuff
is supposed to work, we certainly haven't intentionally caused a problem.
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: [BK] disconnected operation

[James Bottomley]

On Sun, 2004-12-26 at 09:19 -0800, Larry McVoy wrote:
> For James, could you do a little debugging please?  Run the following
> when you are plugged in and it works and also when it doesn't:
> 
> 	bk getuser
> 	bk getuser -r
> 	bk gethost
> 	bk gethost -r
> 	bk dotbk
> 
> We'll track it down and fix it if it is a problem on our end.  This stuff
> is supposed to work, we certainly haven't intentionally caused a problem.

OK, I cloned a new repository and started applying patches to it.  The
transcript of what I did is attached.  You can see that after I
disconnect from the network, I get three emails imported before it spits
an error at me.

James

jejb@mulgrave> bk lease renew
jejb@mulgrave> PATH=/home/jejb/BK/BK-kernel-tools:$PATH
jejb@mulgrave> dotest < ~/tmp.mail
bk import -tpatch -CR -yibmvscsi.c: replace schedule_timeout() with msleep() /tmp/patch20138 .
Patching...
Patching file drivers/scsi/ibmvscsi/ibmvscsi.c
Checking for potential renames in /home/jejb/BK/test-2.6 ...
Checking in new or modified files in /home/jejb/BK/test-2.6 ...
bk commit -y[PATCH] ibmvscsi.c: replace schedule_timeout() with msleep()

Description: Use msleep() instead of schedule_timeout()
to guarantee the task delays as expected.  Originally
submitted to linux-scsi by the janitors, and resubmitted
by boutcher (after testing :-)

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Maximilian Attems <janitor@sternwelten.at>
Signed-off-by: Dave Boutcher <boutcher@us.ibm.com>

ChangeSet revision 1.2206: +1 -0 = 38195
Sending ChangeSet log ...
jejb@mulgrave> bk getuser
jejb
jejb@mulgrave> bk getuser -r
jejb
jejb@mulgrave> bk gethost
mulgrave.(none)
jejb@mulgrave> bk gethost -r
mulgrave.(none)
jejb@mulgrave> bk dotbk
/home/jejb/.bk
jejb@mulgrave> dotest < ~/tmp.mail
bk import -tpatch -CR -yibmvscsi.c: limit size of I/O requests /tmp/patch20470 .
Patching...
Patching file drivers/scsi/ibmvscsi/ibmvscsi.c
Checking for potential renames in /home/jejb/BK/test-2.6 ...
Checking in new or modified files in /home/jejb/BK/test-2.6 ...
bk commit -y[PATCH] ibmvscsi.c: limit size of I/O requests

Description: Limit the size of I/O requests sent by the
ibmvscsi adapter.  With better I/O scheduling (and thus larger
requests) we were breaking some servers.

Signed-off-by: Dave Boutcher <boutcher@us.ibm.com>

ChangeSet revision 1.2207: +1 -0 = 38196
Sending ChangeSet log ...

[HERE I FLIP OUT THE WIRELESS CARD TO DISCONNECT]

jejb@mulgrave> dotest < ~/tmp.mail
bk import -tpatch -CR -yscsi/aic7xxx/aic79xx_osm.c: remove an unused function /tmp/patch20653 .
Patching...
Patching file drivers/scsi/aic7xxx/aic79xx_osm.c
Checking for potential renames in /home/jejb/BK/test-2.6 ...
Checking in new or modified files in /home/jejb/BK/test-2.6 ...
bk commit -y[PATCH] scsi/aic7xxx/aic79xx_osm.c: remove an unused function

[ this time without the problems due to a digital signature... ]

The patch below removes an unused function from
drivers/scsi/aic7xxx/aic79xx_osm.c


diffstat output:
 drivers/scsi/aic7xxx/aic79xx_osm.c |   26 --------------------------
 1 files changed, 26 deletions(-)


Signed-off-by: Adrian Bunk <bunk@stusta.de>

ChangeSet revision 1.2208: +1 -0 = 38197
Sending ChangeSet log ...
jejb@mulgrave> dotest < ~/tmp.mail
bk import -tpatch -CR -yscsi/ahci.c: remove an unused function /tmp/patch20760 .
Patching...
Patching file drivers/scsi/ahci.c
Checking for potential renames in /home/jejb/BK/test-2.6 ...
Checking in new or modified files in /home/jejb/BK/test-2.6 ...
bk commit -y[PATCH] scsi/ahci.c: remove an unused function

[ this time without the problems due to a digital signature... ]

The patch below removes an unused function from drivers/scsi/ahci.c


diffstat output:
 drivers/scsi/ahci.c |    9 ---------
 1 files changed, 9 deletions(-)


Signed-off-by: Adrian Bunk <bunk@stusta.de>

ChangeSet revision 1.2209: +1 -0 = 38198
Sending ChangeSet log ...
jejb@mulgrave> dotest < ~/tmp.mail
bk import -tpatch -CR -ygdth: reduce large on-stack locals /tmp/patch20867 .
Patching...
Patching file drivers/scsi/gdth.c
Patching file drivers/scsi/gdth_proc.c
Checking for potential renames in /home/jejb/BK/test-2.6 ...
Checking in new or modified files in /home/jejb/BK/test-2.6 ...
bk commit -y[PATCH] gdth: reduce large on-stack locals

gdth is the fourth-highest stack user (for a single function)
in 2.6.10-rc3-bk-recent (sizes on x86-32).

Reduce stack usage in gdth driver:
reduce ioc_rescan() from 1564 to 52 bytes;
reduce ioc_hdrlist() from 1528 to 24 bytes;
reduce gdth_get_info() from 1076 to 300 bytes;

Signed-off-by: Randy Dunlap <rddunlap@osdl.org>

diffstat:=
 drivers/scsi/gdth.c      |  194 ++++++++++++++++++++++++++---------------------
 drivers/scsi/gdth_proc.c |  145 ++++++++++++++++++-----------------
 2 files changed, 189 insertions(+), 150 deletions(-)

[HERE BK POPS UP A DIALOGUE SAYING:

Unable to obtain permission to use this version of BitKeeper (bk-3.2.3) 
from lease.openlogging.org.  That server issues certificates to use BK
for openlogging for 30 days at a time.  The bk binary needs to be able
to make a http connection to lease.openlogging.org at least once a month.

Look at 'bk help url' if you need to tell 'bk' about a proxy.

AND THE COMMIT FAILS WITH THE FOLLOWING:]

You need to figure out why you have two files with the same ID
and correct that situation before this ChangeSet can be created.
jejb@mulgrave> bk getuser
jejb
jejb@mulgrave> bk getuser -r
jejb
jejb@mulgrave> bk gethost
mulgrave.(none)
jejb@mulgrave> bk gethost -r
mulgrave.(none)
jejb@mulgrave> bk dotbk
/home/jejb/.bk

Re: [BK] disconnected operation

[Larry McVoy]

> OK, I cloned a new repository and started applying patches to it.  The
> transcript of what I did is attached.  You can see that after I
> disconnect from the network, I get three emails imported before it spits
> an error at me.

OK, cool.  We can keep going back and forth or if you wish you can send me
the mailbox of patches and the cset key to which they should be applied 
and I'll try it.

Can you do a "bk lease renew" before you start the process, then do a
"bk lease show" to make sure it took?  When it starts to fail I'd like
to know what time your computer thinks it is.  Is it possible that you
are using your net connection to maintain your date and then when you
disconnect your date warps forward?  Does this always happen at the 
3rd commit?

Thanks!
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: [BK] disconnected operation

[Zwane Mwaikambo]

On Sun, 26 Dec 2004, Larry McVoy wrote:

> On Sun, Dec 26, 2004 at 10:43:13AM -0600, James Bottomley wrote:
> > On Sun, 2004-12-26 at 08:27 -0800, Larry McVoy wrote:
> > > I suspect that your hostname changes when you disconnect.  Leases are 
> > > issued on a per host basis.  If you make your hostname constant when
> > > you unplug it should work.  If it doesn't, let us know.
> > 
> > Well, that's a new one, but no, I have a fixed hostname which dhcp is
> > forbidden from changing.
> 
> Let's do a little poll here to find out if it is specific to you or if
> this is a problem that everyone is having.  Could we get people who
> use BK disconnected to stand up and be counted?  Does this work for 
> anyone?

I use it whilst on travel too, however i do not have a similar problem as 
described by James but i've noticed that simple operations like `bk vi 
filename` take extremely long;

zwane@r3000 ~ {0:0} bk version
BitKeeper version is bk-3.2.3 20040818155841 for x86-glibc23-linux
Built by: lm@redhat9.bitmover.com in /build/3.2.x-lm/src
Built on: Wed Aug 18 11:18:31 PDT 2004
Running on: Linux 2.6.10-R3000

zwane@r3000 linux {0:0} time bk cat user.h
#include <asm/user.h>
0.032u 0.004s 0:40.15 0.0%      0+0k 0+0io 1pf+0w

zwane@r3000 linux-2.6-bk {1:0} time bk get Makefile
Makefile 1.551: 1318 lines
0.074u 0.014s 0:32.79 0.2%      0+0k 0+0io 0pf+0w

My hostname is fixed, i do not have a default gateway set and none of the 
nameservers listed in /etc/resolv.conf are reachable. This could very well 
be a local configuration problem but you did ask for people who use BK 
disconnected to stand up =)

Thanks,
	Zwane

Re: [BK] disconnected operation

[James Bottomley]

On Sun, 2004-12-26 at 10:35 -0800, Larry McVoy wrote:
> OK, cool.  We can keep going back and forth or if you wish you can send me
> the mailbox of patches and the cset key to which they should be applied 
> and I'll try it.

That might be easier, since BK operations take a large amount of time on
my laptop.  I'll send you the emails under separate cover.

> Can you do a "bk lease renew" before you start the process, then do a
> "bk lease show" to make sure it took?  When it starts to fail I'd like
> to know what time your computer thinks it is.  Is it possible that you
> are using your net connection to maintain your date and then when you
> disconnect your date warps forward?  Does this always happen at the 
> 3rd commit?

I did do a bk lease renew at the top (in the log).  My clock is
controlled by NTP, but it just syncs to the localhost fudge when it
loses all net connection.  The time doesn't jump when this happens (it
does drift by a few hundred milliseconds every hour or so I remain
disconnected, though).

Where it happens seems to be variable.  Most often it's the first or
second commit.

James

Re: [BK] disconnected operation

[Martin Dalecki]

On 2004-12-26, at 18:19, Larry McVoy wrote:
> For James, could you do a little debugging please?  Run the following
> when you are plugged in and it works and also when it doesn't:
>
> 	bk getuser
> 	bk getuser -r
> 	bk gethost
> 	bk gethost -r
> 	bk dotbk
>
> We'll track it down and fix it if it is a problem on our end.  This 
> stuff
> is supposed to work, we certainly haven't intentionally caused a 
> problem.

Larry, that's futile, please trust me, you really can't fix this 
conveniently
and properly, without changing the principles of operation. nscd, 
mDNSresponder,
dhclient, glibc, NDIS, netinfo, some NAT, proxy, OpenDirectory, and so 
on, the whole gang,
gosh even /etc/hostname will *always* get you sometime if you intend to
accommodate mobile users. Eee... did I say mobile users, Well let's use 
the proper
buzz-word. *Roaming* users.  Take it as a given: The nice shiny easy 
days of BSD4-Net
release are long past and gone. A hostname simply isn't a fixed 
attribute of a host anymore.
Setting a host name has nowadays more the character of a prayer or 
christmas wish to Santa,
then any setup measure.

I really recommend that you give him a cookie and check if he still 
didn't eat it
next time you meet him. (So don't give him a crispy well smelling 
one...)
Or even better please do yourself the favor and go straight the whole 
way down to
SASL, TSL and so on... There where already people around there with 
similar problems before.
(www.beepcore.org comes in to mind for something quite modern and not 
directly
Java/XML/SOAP/bla bla bloated...)

Re: [BK] disconnected operation

[Martin Dalecki]

On 2004-12-26, at 19:41, Zwane Mwaikambo wrote:

> On Sun, 26 Dec 2004, Larry McVoy wrote:
>
>> On Sun, Dec 26, 2004 at 10:43:13AM -0600, James Bottomley wrote:
>>> On Sun, 2004-12-26 at 08:27 -0800, Larry McVoy wrote:
>>>> I suspect that your hostname changes when you disconnect.  Leases 
>>>> are
>>>> issued on a per host basis.  If you make your hostname constant when
>>>> you unplug it should work.  If it doesn't, let us know.
>>>
>>> Well, that's a new one, but no, I have a fixed hostname which dhcp is
>>> forbidden from changing.
>>
>> Let's do a little poll here to find out if it is specific to you or if
>> this is a problem that everyone is having.  Could we get people who
>> use BK disconnected to stand up and be counted?  Does this work for
>> anyone?
>
> I use it whilst on travel too, however i do not have a similar problem 
> as
> described by James but i've noticed that simple operations like `bk vi
> filename` take extremely long;

There are very few name-service implementations out there with proper
error handling out there. Sad world this is ;-) ...

Re: [BK] disconnected operation

[Marcelo Tosatti]

On Sun, Dec 26, 2004 at 09:19:00AM -0800, Larry McVoy wrote:
> On Sun, Dec 26, 2004 at 10:43:13AM -0600, James Bottomley wrote:
> > On Sun, 2004-12-26 at 08:27 -0800, Larry McVoy wrote:
> > > I suspect that your hostname changes when you disconnect.  Leases are 
> > > issued on a per host basis.  If you make your hostname constant when
> > > you unplug it should work.  If it doesn't, let us know.
> > 
> > Well, that's a new one, but no, I have a fixed hostname which dhcp is
> > forbidden from changing.
> 
> Let's do a little poll here to find out if it is specific to you or if
> this is a problem that everyone is having.  Could we get people who
> use BK disconnected to stand up and be counted?  Does this work for 
> anyone?
> 
> For James, could you do a little debugging please?  Run the following
> when you are plugged in and it works and also when it doesn't:
> 
> 	bk getuser
> 	bk getuser -r
> 	bk gethost
> 	bk gethost -r
> 	bk dotbk
> 
> We'll track it down and fix it if it is a problem on our end.  This stuff
> is supposed to work, we certainly haven't intentionally caused a problem.

Larry,

I have never been able to use BK in disconnected mode, which is very annoying.

It fails to connect to lease.openlogging.org as James describes.

Is disconnected operation supposed to work ? It didnt seem so.

Re: [BK] disconnected operation

[Ricky Beam]

On Sun, 26 Dec 2004, Martin Dalecki wrote:
>A hostname simply isn't a fixed attribute of a host anymore.

It is on properly setup and maintained machines.

The problem are all those people writing programs that think they are doing
the world a favor by screwing with the hostname and various other settings
for us... there's no reason for dhcp to change my hostname.  At least on
linux, no dhcp implementation touches /etc/hosts. (Solaris has screwed up
the hosts file for years.)

These are the same machines that don't have FQDN's as the first name per
entry in /etc/hosts (which pisses off many incarnations of glibc.) *grin*

--Ricky

PS: When you're off-line, /etc/resolv.conf shouldn't have any nameservers
    listed.  They aren't going to be connectable.

Re: [BK] disconnected operation

[Kyle Moffett]

On Dec 28, 2004, at 09:33, Ricky Beam wrote:
> On Sun, 26 Dec 2004, Martin Dalecki wrote:
>> A hostname simply isn't a fixed attribute of a host anymore.
>
> It is on properly setup and maintained machines.
>
> The problem are all those people writing programs that think they are 
> doing
> the world a favor by screwing with the hostname and various other 
> settings
> for us... there's no reason for dhcp to change my hostname.  At least 
> on
> linux, no dhcp implementation touches /etc/hosts. (Solaris has screwed 
> up
> the hosts file for years.)
>
> These are the same machines that don't have FQDN's as the first name 
> per
> entry in /etc/hosts (which pisses off many incarnations of glibc.) 
> *grin*

So what would happen to somebody who put their BK files on a portable
drive and carried it from home to work.  That's a perfectly reasonable 
thing to
do, both for security and for speed reasons, but it would appear to 
cause
problems.

Cheers,
Kyle Moffett

-----BEGIN GEEK CODE BLOCK-----
Version: 3.12
GCM/CS/IT/U d- s++: a18 C++++>$ UB/L/X/*++++(+)>$ P+++(++++)>$
L++++(+++) E W++(+) N+++(++) o? K? w--- O? M++ V? PS+() PE+(-) Y+
PGP+++ t+(+++) 5 X R? tv-(--) b++++(++) DI+ D+ G e->++++$ h!*()>++$ r  
!y?(-)
------END GEEK CODE BLOCK------

Re: [BK] disconnected operation

[Ricky Beam]

On Wed, 29 Dec 2004, Kyle Moffett wrote:
>So what would happen to somebody who put their BK files on a portable
>drive and carried it from home to work.  That's a perfectly reasonable
>thing to do, both for security and for speed reasons, but it would appear
>to cause problems.

First, the license(s) are stored in the user's home directory (~/.bk/lease)
per hostname.  If you move to a completely different machine, then, yes,
there will need to be a lease for that machine.

What you are describing is no different from the NFS case.  It doesn't
matter that the media has physically moved; it's still visible to multiple,
unique hosts.  Each host(name) will need it's own lease.

--Ricky

Re: lease.openlogging.org is unreachable

[David Brownell]

> There's something in BK that refuses to work when it can't contact
> lease.openlogging.org, regardless of whether you just renewed the lease
> or not.  

For the record, I've seen this regularly too.  BK 3.2.3
was the last that showed it for me.  I'm glad to know
it wasn't just me ... but would be more glad if none
of us saw the problem!  :)

Last time I looked at the failure mode there was no issue
of network connectivity being missing, or even changing.
Host names on these systems don't change; neither do IP
addresses (though they're behind a NAT gateway).

>From memory (== may not be quite right), what I did was
"bk lease renew", then "bk pull" (or maybe clone; it failed
because it thought openlogging.org was playing hide/seek);
then "bk lease show" showed three (!!) current licences.

I've also had the "...unreachable" messages immediately after
renewing a lease, when cloning from a private tree but with
the network link down.  (That is, same as previous setup,
except the link is down.)  It's actually kind of annoying
to fail at the _end_ of a clone operation, leaving a tree
that doesn't seem recoverable, rather than at the start.

- Dave

Re: lease.openlogging.org is unreachable

[Ricky Beam]

On Mon, 27 Dec 2004, Martin Dalecki wrote:
>for example the pure hobby-student-project ClearCase are going to the
>extent of requiring you to run their own networked filesystem if running
>on Linux.

WTF are talking about?  MVFS is required *everywhere* you want to use
dynamic views... solaris, aix, hpux, windows, and *yes* LINUX.

And MVFS is not 100% filesystem -- it's part fs, part db, part proxy, ...
(it's just weird, but I like it.)

--Ricky

Re: lease.openlogging.org is unreachable

[Ricky Beam]

On Mon, 27 Dec 2004, Martin Dalecki wrote:
...
>I don't know of a proper solution, other then writing in big letters
>in the documentation about the circumstances under which you can shoot
>yourself in
>the feet.
>
>Where is NFSv4?

The problem is the protocol, it's the fact that the directory is shared among
more than one machine.  It's the false assumption that a given location is
unique to just one machine.  Yes, I've fallen into this mess too. (And I wasn't
using NFS, btw.)

--Ricky

Re: lease.openlogging.org is unreachable

[Bernd Eckenfels]

In article <20041227024346.GA3595@work.bitmover.com> you wrote:
> And that, my friends, is the difference between a project which is a hobby
> and a project which is a product.  We don't have the option of saying "this
> doesn't work".  

Thats why more and more commercial solutions do not use  licensing services
or dongles anymore. It just introduces to much failure possibilities.

Greetings
Bernd

Re: lease.openlogging.org is unreachable

[Martin Dalecki]

On 2004-12-27, at 03:43, Larry McVoy wrote:

> On Mon, Dec 27, 2004 at 03:40:38AM +0100, Martin Dalecki wrote:
>>
>> I don't know of a proper solution, other then writing in big letters
>> in the documentation about the circumstances under which you can shoot
>> yourself in
>> the feet.
>
> And that, my friends, is the difference between a project which is a 
> hobby
> and a project which is a product.  We don't have the option of saying 
> "this
> doesn't work".

Fully agreed. And those are the reasons why the completely non 
professional
solutions, which I never saw in deployment of course...
like ... for example the pure hobby-student-project ClearCase are going 
to the
extent of requiring you to run their own networked filesystem if running
on Linux. They do it simply out of fun and for educational purposes and
because they where to dumb to find much easier solution which are out 
there ready
and waiting. They simply don't have to deal with something as 
exceptional as
customers. They love to spend about 10 man years to develop a reliable 
networked
filesystem which is OS minor version specific like burning money for 
heat...

OK I was just kidding. ;-)

Re: lease.openlogging.org is unreachable

[Larry McVoy]

On Mon, Dec 27, 2004 at 03:40:38AM +0100, Martin Dalecki wrote:
> Citing from: 
> http://www.time-travellers.org/shane/papers/NFS_considered_harmful.html
> 
> "NFS fails at the goal of allowing a computer to access files over a 
> network as if they were on a local disk. In many ways, NFS comes close 
> to the objective, and in certain circumstances (detailed later), this 
> is acceptable. However, the subtle differences can cause subtle bugs 
> and greater system issues. The widespread misconception about the 
> compatibility and transparency of NFS means that it is often used 
> inappropriately, and often put into production when better, more 
> acceptable solutions exist."
> 
> I don't know of a proper solution, other then writing in big letters
> in the documentation about the circumstances under which you can shoot 
> yourself in
> the feet.

And that, my friends, is the difference between a project which is a hobby
and a project which is a product.  We don't have the option of saying "this
doesn't work".  
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: lease.openlogging.org is unreachable

[Martin Dalecki]

On 2004-12-27, at 02:41, Larry McVoy wrote:

> On Sun, Dec 26, 2004 at 09:31:11PM -0300, Horst von Brand wrote:
>>> The other answer, which I'm happy to consider, is to come up with a 
>>> unique
>>> id on a per host basis and use that for the leases.  That's not a 
>>> fun task,
>>> does anyone have code (BSD license please) which does that?
>>
>> MAC of eth0?
>
> As others have pointed out that won't work.
>
> I'm trying to remember why we get leases on a per host basis and I 
> think
> it is for a simple reason, NFS.

Citing from: 
http://www.time-travellers.org/shane/papers/NFS_considered_harmful.html

"NFS fails at the goal of allowing a computer to access files over a 
network as if they were on a local disk. In many ways, NFS comes close 
to the objective, and in certain circumstances (detailed later), this 
is acceptable. However, the subtle differences can cause subtle bugs 
and greater system issues. The widespread misconception about the 
compatibility and transparency of NFS means that it is often used 
inappropriately, and often put into production when better, more 
acceptable solutions exist."

I don't know of a proper solution, other then writing in big letters
in the documentation about the circumstances under which you can shoot 
yourself in
the feet.

Where is NFSv4?

Re: lease.openlogging.org is unreachable

[Larry McVoy]

On Sun, Dec 26, 2004 at 09:31:11PM -0300, Horst von Brand wrote:
> > The other answer, which I'm happy to consider, is to come up with a unique
> > id on a per host basis and use that for the leases.  That's not a fun task,
> > does anyone have code (BSD license please) which does that?
> 
> MAC of eth0?

As others have pointed out that won't work.

I'm trying to remember why we get leases on a per host basis and I think
it is for a simple reason, NFS.  We update the leases in your home
directory and if your home directory is nfs mounted then we can corrupt
the leases file due to races (yes, we saw this all the time when we had
one leases file).  So we stick the leases for a particular host in that
host's file.
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: lease.openlogging.org is unreachable

[Martin Dalecki]

On 2004-12-27, at 01:31, Horst von Brand wrote:

> Larry McVoy <lm@bitmover.com> said:
>
> [...]
>
>> The other answer, which I'm happy to consider, is to come up with a 
>> unique
>> id on a per host basis and use that for the leases.  That's not a fun 
>> task,
>> does anyone have code (BSD license please) which does that?
>
> MAC of eth0?

ippp0?

Re: lease.openlogging.org is unreachable

[Chris Wedgwood]

On Sun, Dec 26, 2004 at 09:31:11PM -0300, Horst von Brand wrote:

> MAC of eth0?

that changes for people (most commonly laptops)

Re: lease.openlogging.org is unreachable

[Horst von Brand]

Larry McVoy <lm@bitmover.com> said:

[...]

> The other answer, which I'm happy to consider, is to come up with a unique
> id on a per host basis and use that for the leases.  That's not a fun task,
> does anyone have code (BSD license please) which does that?

MAC of eth0?
-- 
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                     Fono: +56 32 654431
Universidad Tecnica Federico Santa Maria              +56 32 654239
Casilla 110-V, Valparaiso, Chile                Fax:  +56 32 797513

Re: lease.openlogging.org is unreachable

[Martin Dalecki]

On 2004-12-27, at 01:03, Chris Wedgwood wrote:

> On Mon, Dec 27, 2004 at 12:13:55AM +0100, Martin Dalecki wrote:
>
>> No the first value that worked in the whole setup of course. And I
>> said it clearly that would be just a workaround because uniqueness
>> can be established in the easiest way on the side of the part which
>> is interested in the persistency - the server peer not the client.
>
> actually, the bk lease server could give out id's and those could be
> caches in ~/.bk<whatever> --- server side it could be a counter that
> you just xor with some s3kr1t value and then blind using a hash or
> cryto-function, something good-enough (statistically unlikely to
> break) is all that is required, it doesn't have to be perfect surely?
>

Right that's precisely what's called a cookie or token. But it would 
involve
a change in the on-wire protocol.

Re: lease.openlogging.org is unreachable

[Chris Wedgwood]

On Mon, Dec 27, 2004 at 12:13:55AM +0100, Martin Dalecki wrote:

> No the first value that worked in the whole setup of course. And I
> said it clearly that would be just a workaround because uniqueness
> can be established in the easiest way on the side of the part which
> is interested in the persistency - the server peer not the client.

actually, the bk lease server could give out id's and those could be
caches in ~/.bk<whatever> --- server side it could be a counter that
you just xor with some s3kr1t value and then blind using a hash or
cryto-function, something good-enough (statistically unlikely to
break) is all that is required, it doesn't have to be perfect surely?

Re: lease.openlogging.org is unreachable

[Martin Dalecki]

On 2004-12-27, at 00:01, Florian Weimer wrote:

> * Martin Dalecki:
>
>> Simply storing the first hostname used in a dot file for subsequent
>> reuse on client side, would be even easier I guess.
>
> You mean localhost.localdomain? 8-)

No the first value that worked in the whole setup of course. And I said 
it clearly that would
be just a workaround because uniqueness can be established in the 
easiest way on the side of
the part which is interested in the persistency - the server peer not 
the client.
"Simple" http servers do it by sending cookies around. Other call it 
"tokens".

Re: lease.openlogging.org is unreachable

[Florian Weimer]

* Martin Dalecki:

> Simply storing the first hostname used in a dot file for subsequent
> reuse on client side, would be even easier I guess.

You mean localhost.localdomain? 8-)

Re: lease.openlogging.org is unreachable

[Martin Dalecki]

On 2004-12-26, at 21:37, Alan Cox wrote:

> On Sul, 2004-12-26 at 18:18, Larry McVoy wrote:
>> The other answer, which I'm happy to consider, is to come up with a 
>> unique
>> id on a per host basis and use that for the leases.  That's not a fun 
>> task,
>> does anyone have code (BSD license please) which does that?
>
> libuuid does that on straight statistical probability - what properties
> do you want your id to have ?
>

Simply storing the first hostname used in a dot file for subsequent 
reuse on client side,
would be even easier I guess. That would be basically the same strategy 
as used by
ssh with regard to host keys. It wouldn't even perhaps make protocol 
changes necessary.
But still not a perfect solution... (Remember the times you have delete 
something from
.ssh/known_hosts).

Re: lease.openlogging.org is unreachable

[Alan Cox]

On Sul, 2004-12-26 at 18:18, Larry McVoy wrote:
> The other answer, which I'm happy to consider, is to come up with a unique
> id on a per host basis and use that for the leases.  That's not a fun task,
> does anyone have code (BSD license please) which does that?

libuuid does that on straight statistical probability - what properties
do you want your id to have ?

Re: lease.openlogging.org is unreachable

[Martin Dalecki]

On 2004-12-26, at 19:18, Larry McVoy wrote:
>
> The other answer, which I'm happy to consider, is to come up with a 
> unique
> id on a per host basis and use that for the leases.  That's not a fun 
> task,
> does anyone have code (BSD license please) which does that?

Nothing even moderately portable like this exists. The best 
approximation
imaginable would be starting to collect MAC address id's... But thats 
of corse
the wrong protocol layer.

Re: lease.openlogging.org is unreachable

[Larry McVoy]

On Sun, Dec 26, 2004 at 07:06:52PM +0100, M?ns Rullg?rd wrote:
> Larry McVoy <lm@bitmover.com> writes:
> 
> > On Sun, Dec 26, 2004 at 11:26:42AM +0100, M?ns Rullg?rd wrote:
> >> I've been bitten by that one, as I occasionally work off-line.  Is
> >> there some way I can make BK renew the leases a week or so before they
> >> expire?
> >
> > In theory, we do that already.  By you can always do a "bk lease renew"
> > and that will get you a new lease.  "bk lease show" will show you your
> > lease.
> 
> Looking closer, the problem is that my hostname keeps changing,
> depending on which network the laptop is on, if any.  I guess the
> simple solution is to remember to renew the lease for the no-net
> hostname before going offline.

The other answer, which I'm happy to consider, is to come up with a unique
id on a per host basis and use that for the leases.  That's not a fun task,
does anyone have code (BSD license please) which does that?
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: lease.openlogging.org is unreachable

[Måns Rullgård]

Larry McVoy <lm@bitmover.com> writes:

> On Sun, Dec 26, 2004 at 11:26:42AM +0100, M?ns Rullg?rd wrote:
>> I've been bitten by that one, as I occasionally work off-line.  Is
>> there some way I can make BK renew the leases a week or so before they
>> expire?
>
> In theory, we do that already.  By you can always do a "bk lease renew"
> and that will get you a new lease.  "bk lease show" will show you your
> lease.

Looking closer, the problem is that my hostname keeps changing,
depending on which network the laptop is on, if any.  I guess the
simple solution is to remember to renew the lease for the no-net
hostname before going offline.

-- 
Måns Rullgård
mru@inprovide.com

Re: lease.openlogging.org is unreachable

[Larry McVoy]

On Sun, Dec 26, 2004 at 11:26:42AM +0100, M?ns Rullg?rd wrote:
> I've been bitten by that one, as I occasionally work off-line.  Is
> there some way I can make BK renew the leases a week or so before they
> expire?

In theory, we do that already.  By you can always do a "bk lease renew"
and that will get you a new lease.  "bk lease show" will show you your
lease.
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: lease.openlogging.org is unreachable

[Chuck Ebbert]

Larry McVoy wrote:

> The interesting thing is that the code already has a backup in it and I just
> checked that code path and it works.

 Huh.  The first time the lease renewal hung up and I hit ctrl-C when it
didn't seem to be working (tcpdump showed a series of unanswered SYNs.)
Later I let it run to completion and it must have worked, but since there
was no output from the program I assumed it had just failed silently like
some other bitkeeper commands will do. Then the Christmas morning mayhem
hit and I didn't have any more time to play with it.

 You should output a confirmation message when 'bk lease renew' succeeds.

--
Please take it as a sign of my infinite respect for you,
that I insist on you doing all the work.
                                        -- Rusty Russell

Re: lease.openlogging.org is unreachable

[Måns Rullgård]

Larry McVoy <lm@bitmover.com> writes:

> The interesting thing is that the code already has a backup in it and I just
> checked that code path and it works.
>
> Has anyone else been shut down because of lease.openlogging.org being down
> and if so what version of BK were you running please?
>
> It is true that both servers are at our offices so if the network had been
> down you would have been out of luck.

I've been bitten by that one, as I occasionally work off-line.  Is
there some way I can make BK renew the leases a week or so before they
expire?

It would also be nice if a simple read-only 'get' were allowed without
a lease at all.

-- 
Måns Rullgård
mru@inprovide.com

Re: lease.openlogging.org is unreachable

[Larry McVoy]

The interesting thing is that the code already has a backup in it and I just
checked that code path and it works.

Has anyone else been shut down because of lease.openlogging.org being down
and if so what version of BK were you running please?

It is true that both servers are at our offices so if the network had been
down you would have been out of luck.  We'll create some more backups and
scatter them around the US, that's no problem.  But if the logic in the 
code to get them is busted then we need to fix that.

Things should be stabilized now, the machine went up and down a few times
after my last "it's OK mail", it turned out we had a dieing ethernet card;
it's been replaced and we seem stable now.  
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: lease.openlogging.org is unreachable

[Larry McVoy]

On Sat, Dec 25, 2004 at 01:20:34AM -0500, Chuck Ebbert wrote:
> lease.openlogging.org is unreachable today.
> 
> So I guess I need to set up a cron job to renew my lease every
> minute/hour/day/whatever so I can actually download new kernel
> releases when they come out?  I can't even examine the code I
> downloaded yesterday without that lease...  Now that's what I call
> having my source code held hostage!

It's back up.  Go to a repo and say "bk lease renew" and you should be
all set.  
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: lease.openlogging.org is unreachable

[Larry McVoy]

On Sat, Dec 25, 2004 at 12:24:48PM +0000, Alan Cox wrote:
> On Sad, 2004-12-25 at 06:20, Chuck Ebbert wrote:
> > lease.openlogging.org is unreachable today.
> > 
> > So I guess I need to set up a cron job to renew my lease every
> > minute/hour/day/whatever so I can actually download new kernel
> > releases when they come out?  I can't even examine the code I
> > downloaded yesterday without that lease...  Now that's what I call
> > having my source code held hostage!
> 
> The tar ball edition works perfectly 8)

And a very Merry Christmas to you too Alan.  You're welcome for all that
BK has done for you and the kernel effort, always a pleasure to work with
such polite people.
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: lease.openlogging.org is unreachable

[Larry McVoy]

On Sat, Dec 25, 2004 at 01:20:34AM -0500, Chuck Ebbert wrote:
> lease.openlogging.org is unreachable today.

Yup it is, I'm trying to reboot it remotely but it doesn't seem to want
to come up.  I'll go in and figure it out, it will be about 30 minutes 
before I can get there though.  

> So I guess I need to set up a cron job to renew my lease every
> minute/hour/day/whatever so I can actually download new kernel
> releases when they come out?  I can't even examine the code I
> downloaded yesterday without that lease...  Now that's what I call
> having my source code held hostage!

We'll put an offsite backup in place so it will take two failures to
cause you a problem.

In general the leases are good for 30 days so only a few people should
be effected by this.  
-- 
---
Larry McVoy                lm at bitmover.com           http://www.bitkeeper.com

Re: lease.openlogging.org is unreachable

[Alan Cox]

On Sad, 2004-12-25 at 06:20, Chuck Ebbert wrote:
> lease.openlogging.org is unreachable today.
> 
> So I guess I need to set up a cron job to renew my lease every
> minute/hour/day/whatever so I can actually download new kernel
> releases when they come out?  I can't even examine the code I
> downloaded yesterday without that lease...  Now that's what I call
> having my source code held hostage!

The tar ball edition works perfectly 8)

lease.openlogging.org is unreachable

[Chuck Ebbert]

lease.openlogging.org is unreachable today.

So I guess I need to set up a cron job to renew my lease every
minute/hour/day/whatever so I can actually download new kernel
releases when they come out?  I can't even examine the code I
downloaded yesterday without that lease...  Now that's what I call
having my source code held hostage!

--
Please take it as a sign of my infinite respect for you,
that I insist on you doing all the work.
                                        -- Rusty Russell