[gentoo-user] Server system date synchronizaion

[gentoo-user] Server system date synchronizaion

[Nick Khamis]

Hello Everyone,

We are trying to sync our server's time with an accurate ntp
server, and was wondering which of the many solutions are
considered viable. I did see the
http://en.gentoo-wiki.com/wiki/Time_Synchronization.
Our services are quite time sensitive.

Thanks in Advance,

N.

Re: [gentoo-user] Server system date synchronizaion

[Dale]

Nick Khamis wrote:
> Hello Everyone,
>
> We are trying to sync our server's time with an accurate ntp
> server, and was wondering which of the many solutions are
> considered viable. I did see the
> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
> Our services are quite time sensitive.
>
> Thanks in Advance,
>
> N.
>
>

net-misc/ntp
net-misc/openntpd
net-misc/chrony

One of those should work.  I think the plain ntp has been around the
longest.  I couldn't get it to work right on my rig so I switched to
chrony.  Basically, I would try ntp first then go from there if needed.

Hope that helps.

Dale

:-)  :-) 

-- 
I am only responsible for what I said ... Not for what you understood or how you interpreted my words!

Re: [gentoo-user] Server system date synchronizaion

[Michael Mol]

[-- Attachment #1: Type: text/plain, Size: 794 bytes --]

On 04/25/2013 10:33 AM, Nick Khamis wrote:
> Hello Everyone,
> 
> We are trying to sync our server's time with an accurate ntp
> server, and was wondering which of the many solutions are
> considered viable. I did see the
> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
> Our services are quite time sensitive.

My best results so far have been to have one node on my network sync to
pool.ntp.org, and to have all other nodes on my network sync to that one
node. Short of having a stratum 1 time server on my network, that seems
to work the best; done that way, my nodes are within a few milliseconds
of each other, near as I can figure.

For contrast, having all nodes sync to pool.ntp.org results in time
variance of up to 2-3 minutes across a dozen or so machines.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

Re: [gentoo-user] Server system date synchronizaion

[Tanstaafl]

On 2013-04-25 10:40 AM, Michael Mol <mikemol@gmail.com> wrote:
> For contrast, having all nodes sync to pool.ntp.org results in time
> variance of up to 2-3 minutes across a dozen or so machines.

That makes no sense...

Not calling you a liar or anything, but it just doesn't make sense.

I can see that it might take each system different times to get fully 
sync'd, but for them to consistently vary by this amount? No, something 
else is wrong.

Are these virtualized servers?

Re: [gentoo-user] Server system date synchronizaion

[Tanstaafl]

On 2013-04-25 10:33 AM, Nick Khamis <symack@gmail.com> wrote:
> We are trying to sync our server's time with an accurate ntp
> server, and was wondering which of the many solutions are
> considered viable. I did see the
> http://en.gentoo-wiki.com/wiki/Time_Synchronization.

Are these virtualized? It makes a difference, and from everything I've 
read, you don't sync virtualized servers the same as bare metal servers.

> Our services are quite time sensitive.

Ummm... *all* servers are critically time-sensitive.

Re: [gentoo-user] Server system date synchronizaion

[Nick Khamis]

On 4/25/13, Michael Mol <mikemol@gmail.com> wrote:
> On 04/25/2013 10:33 AM, Nick Khamis wrote:
>> Hello Everyone,
>>
>> We are trying to sync our server's time with an accurate ntp
>> server, and was wondering which of the many solutions are
>> considered viable. I did see the
>> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
>> Our services are quite time sensitive.
>
> My best results so far have been to have one node on my network sync to
> pool.ntp.org, and to have all other nodes on my network sync to that one
> node. Short of having a stratum 1 time server on my network, that seems
> to work the best; done that way, my nodes are within a few milliseconds
> of each other, near as I can figure.
>
> For contrast, having all nodes sync to pool.ntp.org results in time
> variance of up to 2-3 minutes across a dozen or so machines.
>
>


Thank you so much for your response. Michael, were you using ntp to
sync that initial server? If so, can we get that setup up and running
easily? I've been putting the time issue
off for way too long...

Thanks in Advance,

Nick

Re: [gentoo-user] Server system date synchronizaion

[Nick Khamis]

> Ummm... *all* servers are critically time-sensitive.
>

Yeah... I concur ;)

Re: [gentoo-user] Server system date synchronizaion

[Michael Mol]

[-- Attachment #1: Type: text/plain, Size: 1938 bytes --]

On 04/25/2013 10:46 AM, Tanstaafl wrote:
> On 2013-04-25 10:40 AM, Michael Mol <mikemol@gmail.com> wrote:
>> For contrast, having all nodes sync to pool.ntp.org results in time
>> variance of up to 2-3 minutes across a dozen or so machines.
> 
> That makes no sense...
> 
> Not calling you a liar or anything, but it just doesn't make sense.
> 
> I can see that it might take each system different times to get fully
> sync'd, but for them to consistently vary by this amount? No, something
> else is wrong.
> 
> Are these virtualized servers?

Some are virtualized, some are hosts, some are standalone.

When all machines were configured to speak to pool.ntp.org, the variance
was high. Obviously more so any time a guest was using its host's clock,
and both guest and host were trying to adjust.

There was still significant difference even between standalone systems.
pool.ntp.org pulls from a huge pool of timeservers, and there is visible
variance between more than a few of them. It's a volunteer effort.
*shrug* Unfortunately, I don't have the exact variances in my notes.

When I used a single standalone to connect to pool.ntp.org, and had all
other systems (standalone, virtualized and guest) connect to that
standalone system, virtually all variance went away. The stability of
having a single local time source for all but one local machine to sync
against overcame the instability caused by having host and guest ntp
clients stacked.


Of course, ideally, you want VM guests to rely on the VM host for their
clock, and have the VM host configured with a good time source. And you
would want all bare iron configured to talk to a small pool of tightly
synchronized time servers. And if you can trust your layer 2 (or secure
your layer 3 with, e.g. ipsec), you may further benefit from setting up
a multicast time source.

Further, ideally, you want a stratum 1 time server locally.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

Re: [gentoo-user] Server system date synchronizaion

[Michael Mol]

[-- Attachment #1: Type: text/plain, Size: 1080 bytes --]

On 04/25/2013 11:02 AM, Tanstaafl wrote:
> On 2013-04-25 10:33 AM, Nick Khamis <symack@gmail.com> wrote:
>> We are trying to sync our server's time with an accurate ntp
>> server, and was wondering which of the many solutions are
>> considered viable. I did see the
>> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
> 
> Are these virtualized? It makes a difference, and from everything I've
> read, you don't sync virtualized servers the same as bare metal servers.
> 
>> Our services are quite time sensitive.
> 
> Ummm... *all* servers are critically time-sensitive.
> 

Some are more critical than others. If you're primarily worried about
kerberos, variance of up to a couple minutes will likely go unnoticed.
If you're dumping logs into splunk, and need second-precision timestamps
to be comparable to each other across a multi-campus network, that's a
different degree of time-sensitive. If you're using a distributed
filesystem with time-sensitive conflict resolution algorithms, you could
easily start caring down to sub-millisecond ranges.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 555 bytes --]

Re: [gentoo-user] Server system date synchronizaion

[Nilesh Govindrajan]

[-- Attachment #1: Type: text/plain, Size: 764 bytes --]

On Thursday 25 April 2013 08:09 PM, Dale wrote:
> Nick Khamis wrote:
>> Hello Everyone,
>>
>> We are trying to sync our server's time with an accurate ntp
>> server, and was wondering which of the many solutions are
>> considered viable. I did see the
>> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
>> Our services are quite time sensitive.
>>
>> Thanks in Advance,
>>
>> N.
>>
>>
> net-misc/ntp
> net-misc/openntpd
> net-misc/chrony
>
> One of those should work.  I think the plain ntp has been around the
> longest.  I couldn't get it to work right on my rig so I switched to
> chrony.  Basically, I would try ntp first then go from there if needed.
>
> Hope that helps.
>
> Dale
>
> :-)  :-)
>

You forgot busybox-ntpd


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4210 bytes --]

Re: [gentoo-user] Server system date synchronizaion

[William Kenworthy]

On 25/04/13 23:07, Nick Khamis wrote:
>> Ummm... *all* servers are critically time-sensitive.
>>
> 
> Yeah... I concur ;)
> 

Define critical! - to my mind if its critical you should be running your
own atomic clock, and something like a pps system to distribute it ...

or somewhere in the middle a local gps receiver for time lock.

Or do you mean reasonably accurate, but closely synced local systems?

My interest after having a stable ntp based hierarchy for years is in
trying to get the same using a cisco router and VMs' - not easy so far!
 When I used an ancient netgear adsl, and a linux firewall/ntp server it
was very good, now ...

Does anyone know a good guide to using time sync in VM's, for both
windows and linux (gentoo) guests using libvirt?  Especially for guests
that are resumed, or the whole virtualisation system is hibernated? (ntp
refuses to resync after guest pause/save/restore/resume (known problem),
even with "tinker panic 0"

My current setup is complicated by using a cisco router (adsl) as the
localnet master via local (ISP/University) time servers - its rather
inaccurate so while the machines are often "locked", its in rather
relative terms :)



ghost#sh ntp ass

      address         ref clock     st  when  poll reach  delay  offset
   disp
+~130.95.128.36    210.9.192.50      2    44    64  377    11.9  -844.4
  213.6
+~116.66.162.4     130.234.255.83    2     8    64  377    48.7  -907.5
  213.3
+~203.0.178.191    43.128.117.84     2    23    64  377    12.2  -891.0
  213.3
 ~192.168.48.1     134.115.4.33      3  9h39    64    0    17.3  -616.8
 16000.
*~27.54.95.11      218.100.43.70     2    42    64  377    12.7  -846.7
  221.4
+~202.127.210.36   223.255.185.2     2    31    64  377    62.2  -845.3
  211.2
+~130.102.128.23   132.163.4.101     2    38    64  377    77.3  -850.4
  212.4
 * master (synced), # master (unsynced), + selected, - candidate, ~
configured
ghost#

asterisk ~ # ntpq -p
     remote           refid      st t when poll reach   delay   offset
jitter
==============================================================================
 ghost.lan.local 27.54.95.11      3 u   64   64  377    1.386  2838.19
513.843
asterisk

Re: [gentoo-user] Server system date synchronizaion

[Dale]

Nilesh Govindrajan wrote:
> On Thursday 25 April 2013 08:09 PM, Dale wrote:
>> Nick Khamis wrote:
>>> Hello Everyone,
>>>
>>> We are trying to sync our server's time with an accurate ntp
>>> server, and was wondering which of the many solutions are
>>> considered viable. I did see the
>>> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
>>> Our services are quite time sensitive.
>>>
>>> Thanks in Advance,
>>>
>>> N.
>>>
>>>
>> net-misc/ntp
>> net-misc/openntpd
>> net-misc/chrony
>>
>> One of those should work.  I think the plain ntp has been around the
>> longest.  I couldn't get it to work right on my rig so I switched to
>> chrony.  Basically, I would try ntp first then go from there if needed.
>>
>> Hope that helps.
>>
>> Dale
>>
>> :-)  :-)
>>
>
> You forgot busybox-ntpd
>

Didn't forget, didn't know about it. ;-)   I just listed the ones I have
heard of and either tried or was told about. 

Let's see if I can remember it for next time tho.  :-)

Dale 

:-)  :-) 

-- 
I am only responsible for what I said ... Not for what you understood or how you interpreted my words!

Re: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 26/04/2013 01:42, William Kenworthy wrote:
> Does anyone know a good guide to using time sync in VM's, for both
> windows and linux (gentoo) guests using libvirt?  Especially for guests
> that are resumed, or the whole virtualisation system is hibernated? (ntp
> refuses to resync after guest pause/save/restore/resume (known problem),
> even with "tinker panic 0"


That's not a bug, it's by design.

If ntpd detects the clock is out by more than X seconds [1], it will not
try to correct the difference, concluding that something is wrong and a
human must decide. It can't easily tell the difference between a resumed
guest (or even that it was resumed at all) and a severe problem.

We fixed this by taking the easy route of least resistance;

1. run ntpdate on startup/restart once before ntpd starts
2. start ntpd as normal
3. a colleague wrote a $MAGIC_HOOK to detect resumed guests that runs
ntpdate once

True, it's a brutal solution and uses a baseball bat where some finesse
might be less ugly, but it suits our needs just fine.

[1] I forget what X is and am too lazy to look it up. Is it 30 seconds
or thereabouts?


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Server system date synchronizaion

[staticsafe]

On 4/25/2013 19:50, Alan McKinnon wrote:
> On 26/04/2013 01:42, William Kenworthy wrote:
>> Does anyone know a good guide to using time sync in VM's, for both
>> windows and linux (gentoo) guests using libvirt?  Especially for guests
>> that are resumed, or the whole virtualisation system is hibernated? (ntp
>> refuses to resync after guest pause/save/restore/resume (known problem),
>> even with "tinker panic 0"
> 
> 
> That's not a bug, it's by design.
> 
> If ntpd detects the clock is out by more than X seconds [1], it will not
> try to correct the difference, concluding that something is wrong and a
> human must decide. It can't easily tell the difference between a resumed
> guest (or even that it was resumed at all) and a severe problem.
> 
> We fixed this by taking the easy route of least resistance;
> 
> 1. run ntpdate on startup/restart once before ntpd starts
> 2. start ntpd as normal
> 3. a colleague wrote a $MAGIC_HOOK to detect resumed guests that runs
> ntpdate once
> 
> True, it's a brutal solution and uses a baseball bat where some finesse
> might be less ugly, but it suits our needs just fine.
> 
> [1] I forget what X is and am too lazy to look it up. Is it 30 seconds
> or thereabouts?
> 
> 

"When first started, the daemon normally polls the servers listed in the
configuration file at 64-s intervals. In order to allow a sufficient
number of samples for the NTP algorithms to reliably discriminate
between correctly operating servers and possible intruders, at least
four valid messages from the majority of servers and peers listed in the
configuration file is required before the daemon can set the local
clock. However, if the difference between the client time and server
time is greater than the panic threshold, which defaults to 1000 s, the
daemon will send a message to the system log and shut down without
setting the clock." [0]

[0] - http://doc.ntp.org/4.1.1/debug.htm
-- 
staticsafe
O< ascii ribbon campaign - stop html mail - www.asciiribbon.org
Please don't top post - http://goo.gl/YrmAb
Don't CC me! I'm subscribed to whatever list I just posted on.

Re: [gentoo-user] Server system date synchronizaion

[William Kenworthy]

On 26/04/13 07:57, staticsafe wrote:
> On 4/25/2013 19:50, Alan McKinnon wrote:
>> On 26/04/2013 01:42, William Kenworthy wrote:
>>> Does anyone know a good guide to using time sync in VM's, for both
>>> windows and linux (gentoo) guests using libvirt?  Especially for guests
>>> that are resumed, or the whole virtualisation system is hibernated? (ntp
>>> refuses to resync after guest pause/save/restore/resume (known problem),
>>> even with "tinker panic 0"
>>
>>
>> That's not a bug, it's by design.
>>
>> If ntpd detects the clock is out by more than X seconds [1], it will not
>> try to correct the difference, concluding that something is wrong and a
>> human must decide. It can't easily tell the difference between a resumed
>> guest (or even that it was resumed at all) and a severe problem.
>>
>> We fixed this by taking the easy route of least resistance;
>>
>> 1. run ntpdate on startup/restart once before ntpd starts
>> 2. start ntpd as normal
>> 3. a colleague wrote a $MAGIC_HOOK to detect resumed guests that runs
>> ntpdate once
>>
>> True, it's a brutal solution and uses a baseball bat where some finesse
>> might be less ugly, but it suits our needs just fine.
>>
>> [1] I forget what X is and am too lazy to look it up. Is it 30 seconds
>> or thereabouts?
>>
>>
> 
> "When first started, the daemon normally polls the servers listed in the
> configuration file at 64-s intervals. In order to allow a sufficient
> number of samples for the NTP algorithms to reliably discriminate
> between correctly operating servers and possible intruders, at least
> four valid messages from the majority of servers and peers listed in the
> configuration file is required before the daemon can set the local
> clock. However, if the difference between the client time and server
> time is greater than the panic threshold, which defaults to 1000 s, the
> daemon will send a message to the system log and shut down without
> setting the clock." [0]
> 
> [0] - http://doc.ntp.org/4.1.1/debug.htm
> 


Keep reading :)

Check out "tinker panic o" I mentioned, or the -g argument to ntpd

The docs say its a "once only" adjustment in one place, but I am not
sure thats actually the case.

BillK

Re: [gentoo-user] Server system date synchronizaion

[Joseph]

On 04/25/13 10:33, Nick Khamis wrote:
>Hello Everyone,
>
>We are trying to sync our server's time with an accurate ntp
>server, and was wondering which of the many solutions are
>considered viable. I did see the
>http://en.gentoo-wiki.com/wiki/Time_Synchronization.
>Our services are quite time sensitive.
>
>Thanks in Advance,
>
>N.

put this script on a cron and enjoy :-)

#!/bin/sh
/usr/bin/rdate -s 128.138.140.44
/sbin/hwclock --systohc

-- 
Joseph

Re: [gentoo-user] Server system date synchronizaion

[Jarry]

On 26-Apr-13 16:10, Joseph wrote:
> On 04/25/13 10:33, Nick Khamis wrote:
>>
>> We are trying to sync our server's time with an accurate ntp
>> server, and was wondering which of the many solutions are
>> considered viable. I did see the
>> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
>> Our services are quite time sensitive.
>
> put this script on a cron and enjoy :-)
>
> #!/bin/sh
> /usr/bin/rdate -s 128.138.140.44
> /sbin/hwclock --systohc

Yeah, enjoy mysterious crashes of some services which die
whenever system time changes rapidly, in one big step
(i.e. dovecot, TS, etc)!

Man, I sincerely hope you do *NOT* mean this seriously.
It might work on desktop but that's definitely NOT the way
time on servers should be updated! Some services are so
sensitive they crash even if you shift time 0.2s back
or forth!

I had even to include "tinker step 0" in my ntpd.conf
just because of that problem (it means ntpd will now never
adjust time by stepping, always only by slewing, which in
my case is max 0.5ms per second)...

Jarry
-- 
_______________________________________________________________
This mailbox accepts e-mails only from selected mailing-lists!
Everything else is considered to be spam and therefore deleted.

Re: [gentoo-user] Server system date synchronizaion

[Nick Khamis]

On 4/26/13, Jarry <mr.jarry@gmail.com> wrote:
> On 26-Apr-13 16:10, Joseph wrote:
>> On 04/25/13 10:33, Nick Khamis wrote:
>>>
>>> We are trying to sync our server's time with an accurate ntp
>>> server, and was wondering which of the many solutions are
>>> considered viable. I did see the
>>> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
>>> Our services are quite time sensitive.
>>
>> put this script on a cron and enjoy :-)
>>
>> #!/bin/sh
>> /usr/bin/rdate -s 128.138.140.44
>> /sbin/hwclock --systohc
>
> Yeah, enjoy mysterious crashes of some services which die
> whenever system time changes rapidly, in one big step
> (i.e. dovecot, TS, etc)!
>
> Man, I sincerely hope you do *NOT* mean this seriously.
> It might work on desktop but that's definitely NOT the way
> time on servers should be updated! Some services are so
> sensitive they crash even if you shift time 0.2s back
> or forth!
>
> I had even to include "tinker step 0" in my ntpd.conf
> just because of that problem (it means ntpd will now never
> adjust time by stepping, always only by slewing, which in
> my case is max 0.5ms per second)...
>
> Jarry
> --
> _______________________________________________________________
> This mailbox accepts e-mails only from selected mailing-lists!
> Everything else is considered to be spam and therefore deleted.
>
>

Hello Everyone,

Thank you for the many solutions however, I am totally lost as to which would
be most reliable in a collocation setting vs. office desktop. What we would like
is to set up our own ntp server which other servers and desktops in our office
syncs to. Is this advised? If so, is there a nice tutorial online?

Kind Regards,

N.

Re: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 26/04/2013 17:27, Nick Khamis wrote:
> Hello Everyone,
> 
> Thank you for the many solutions however, I am totally lost as to which would
> be most reliable in a collocation setting vs. office desktop. What we would like
> is to set up our own ntp server which other servers and desktops in our office
> syncs to. Is this advised? If so, is there a nice tutorial online?

The subject of time is vastly more complex than anyone ever thinks at
first look. Time servers are tiered and are themselves both clients and
servers...

So here's what you do: sync everything to your ISP's time servers.
Chances are good they do a better job than you can, just like with DNS
caching.

When you know more about the subject than you do now, you can venture
into rolling your own. I'm not being rude or funny - time servers are
just one of those things that unless you have special needs and LOTS of
cash, it is so much easier to just let someone else do all the heavy
lifting.


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Server system date synchronizaion

[Nick Khamis]

On 4/26/13, Alan McKinnon <alan.mckinnon@gmail.com> wrote:
> On 26/04/2013 17:27, Nick Khamis wrote:
>> Hello Everyone,
>>
>> Thank you for the many solutions however, I am totally lost as to which
>> would
>> be most reliable in a collocation setting vs. office desktop. What we
>> would like
>> is to set up our own ntp server which other servers and desktops in our
>> office
>> syncs to. Is this advised? If so, is there a nice tutorial online?
>
> The subject of time is vastly more complex than anyone ever thinks at
> first look. Time servers are tiered and are themselves both clients and
> servers...
>
> So here's what you do: sync everything to your ISP's time servers.
> Chances are good they do a better job than you can, just like with DNS
> caching.
>
> When you know more about the subject than you do now, you can venture
> into rolling your own. I'm not being rude or funny - time servers are
> just one of those things that unless you have special needs and LOTS of
> cash, it is so much easier to just let someone else do all the heavy
> lifting.
>
>
> --
> Alan McKinnon
> alan.mckinnon@gmail.com
>
>
>

Hello Alan,

Thank you so much for your response, and I totally understand the
effort vs. benefit challenge. However, is it really that much
trouble/unstable to setup our own ntp
server that syncs with our local isp, and have our internal network sync on it?

N.

Re: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 26/04/2013 17:54, Nick Khamis wrote:
> On 4/26/13, Alan McKinnon <alan.mckinnon@gmail.com> wrote:
>> On 26/04/2013 17:27, Nick Khamis wrote:
>>> Hello Everyone,
>>>
>>> Thank you for the many solutions however, I am totally lost as to which
>>> would
>>> be most reliable in a collocation setting vs. office desktop. What we
>>> would like
>>> is to set up our own ntp server which other servers and desktops in our
>>> office
>>> syncs to. Is this advised? If so, is there a nice tutorial online?
>>
>> The subject of time is vastly more complex than anyone ever thinks at
>> first look. Time servers are tiered and are themselves both clients and
>> servers...
>>
>> So here's what you do: sync everything to your ISP's time servers.
>> Chances are good they do a better job than you can, just like with DNS
>> caching.
>>
>> When you know more about the subject than you do now, you can venture
>> into rolling your own. I'm not being rude or funny - time servers are
>> just one of those things that unless you have special needs and LOTS of
>> cash, it is so much easier to just let someone else do all the heavy
>> lifting.
>>
>>
>> --
>> Alan McKinnon
>> alan.mckinnon@gmail.com
>>
>>
>>
> 
> Hello Alan,
> 
> Thank you so much for your response, and I totally understand the
> effort vs. benefit challenge. However, is it really that much
> trouble/unstable to setup our own ntp
> server that syncs with our local isp, and have our internal network sync on it?


No, it's not THAT much effort. You can get by with installing ntpd on a
single machine, pointing it at the upstream time server and pointing all
your clients to it. It's clearly recorded in the config file, you can't
go wrong.

It's understanding how this weird thing called time works that is the
issue. Take for example leap seconds..... urggggggggggg...

The basic question I suppose is why do you want to do it this way? What
do you feel you will gain by doing it yourself?


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Server system date synchronizaion

[Nick Khamis]

On 4/26/13, Alan McKinnon <alan.mckinnon@gmail.com> wrote:
> On 26/04/2013 17:54, Nick Khamis wrote:
>> On 4/26/13, Alan McKinnon <alan.mckinnon@gmail.com> wrote:
>>> On 26/04/2013 17:27, Nick Khamis wrote:
>>>> Hello Everyone,
>>>>
>>>> Thank you for the many solutions however, I am totally lost as to which
>>>> would
>>>> be most reliable in a collocation setting vs. office desktop. What we
>>>> would like
>>>> is to set up our own ntp server which other servers and desktops in our
>>>> office
>>>> syncs to. Is this advised? If so, is there a nice tutorial online?
>>>
>>> The subject of time is vastly more complex than anyone ever thinks at
>>> first look. Time servers are tiered and are themselves both clients and
>>> servers...
>>>
>>> So here's what you do: sync everything to your ISP's time servers.
>>> Chances are good they do a better job than you can, just like with DNS
>>> caching.
>>>
>>> When you know more about the subject than you do now, you can venture
>>> into rolling your own. I'm not being rude or funny - time servers are
>>> just one of those things that unless you have special needs and LOTS of
>>> cash, it is so much easier to just let someone else do all the heavy
>>> lifting.
>>>
>>>
>>> --
>>> Alan McKinnon
>>> alan.mckinnon@gmail.com
>>>
>>>
>>>
>>
>> Hello Alan,
>>
>> Thank you so much for your response, and I totally understand the
>> effort vs. benefit challenge. However, is it really that much
>> trouble/unstable to setup our own ntp
>> server that syncs with our local isp, and have our internal network sync
>> on it?
>
>
> No, it's not THAT much effort. You can get by with installing ntpd on a
> single machine, pointing it at the upstream time server and pointing all
> your clients to it. It's clearly recorded in the config file, you can't
> go wrong.
>
> It's understanding how this weird thing called time works that is the
> issue. Take for example leap seconds..... urggggggggggg...
>
> The basic question I suppose is why do you want to do it this way? What
> do you feel you will gain by doing it yourself?
>
>
> --
> Alan McKinnon
> alan.mckinnon@gmail.com
>
>
>

Hello Alan,

Thank you so much for your time. Our voip cluster time always vary for
some reason....
And with long distance, that could mean upwards to a dollar a call.

N.

Re: [gentoo-user] Server system date synchronizaion

[Stroller]

On 26 April 2013, at 16:41, Alan McKinnon wrote:
> ...
> So here's what you do: sync everything to your ISP's time servers.
> Chances are good they do a better job than you can, just like with DNS
> caching.


I'm not sure if my ISP offers time servers, but Apple and MS both run time servers which are publicly accessible (presumably from any o/s).

I've never changed my laptop from its default, to sync with time.euro.apple.com, but my Linux boxes all use the public ntp pool, so I was surprised to read the other comments claiming the latter to be inaccurate.

Whenever I restart /etc/init.d/ntpd on my Linux boxes I can see their time match that of my laptop, as consistent as I can see, i.e. less than a second's difference between them.

Stroller.

Re: [gentoo-user] Server system date synchronizaion

[Paul Hartman]

On Thu, Apr 25, 2013 at 9:33 AM, Nick Khamis <symack@gmail.com> wrote:
> Hello Everyone,
>
> We are trying to sync our server's time with an accurate ntp
> server, and was wondering which of the many solutions are
> considered viable. I did see the
> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
> Our services are quite time sensitive.

I think the classic method is to use net-misc/ntp

See the extensive article at http://en.gentoo-wiki.com/wiki/NTP for
great examples and description.

Re: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 26/04/2013 19:11, Nick Khamis wrote:
>>> >> Thank you so much for your response, and I totally understand the
>>> >> effort vs. benefit challenge. However, is it really that much
>>> >> trouble/unstable to setup our own ntp
>>> >> server that syncs with our local isp, and have our internal network sync
>>> >> on it?
>> >
>> >
>> > No, it's not THAT much effort. You can get by with installing ntpd on a
>> > single machine, pointing it at the upstream time server and pointing all
>> > your clients to it. It's clearly recorded in the config file, you can't
>> > go wrong.
>> >
>> > It's understanding how this weird thing called time works that is the
>> > issue. Take for example leap seconds..... urggggggggggg...
>> >
>> > The basic question I suppose is why do you want to do it this way? What
>> > do you feel you will gain by doing it yourself?
>> >
>> >
>> > --
>> > Alan McKinnon
>> > alan.mckinnon@gmail.com
>> >
>> >
>> >
> Hello Alan,
> 
> Thank you so much for your time. Our voip cluster time always vary for
> some reason....
> And with long distance, that could mean upwards to a dollar a call.


Ah, OK. That changes things quite a bit. I have a little bit of
experience with that - I work for a large ISP, we have a large VOIP
department and we run a stratum 2 time server that serves most of the
country.

First things first: you can't just stick any old upstream ntp server in
your config and walk away. You are then reliant on the quality of that
upstream, and far too often other time servers operate on a "good
enough" policy - if it's accurate to about a second, it's good enough
(and for desktop users i.e. most ISP clients, it is good enough).

I don't know how big your operation is, if you have budget I suggest you
invest in a proper master time source that is GPS-driven. We have a
Symmetricom (http://www.symmetricom.com) but it's a mature market with
several vendors. Shop around, prices are less than you'd expect (about
the same as a decent mid-range server and much less than Cisco's routers...)

Weather can get in the way, so back up the device with a decent second
upstream. I have a good one available run by the Science and Technology
Research part of the Dept of Trade and Industry and the third option is
all the other big ISPs around.

Depending on your accuracy needs you could get away without the GPS unit
and just use a good upstream, but I'd fight for the budget for it - tell
management it puts control of billing back in your hands, they always
fall for that one :-)

So the summary would be that I reckon ntpd will do what you want as long
as you chose good reliable time sources. With that in hand, the config
is easy as rather well documented. Shout here ont he list if you need a
hand with this when you come to deployment time




-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 26/04/2013 20:36, Stroller wrote:
> 
> On 26 April 2013, at 16:41, Alan McKinnon wrote:
>> ...
>> So here's what you do: sync everything to your ISP's time servers.
>> Chances are good they do a better job than you can, just like with DNS
>> caching.
> 
> 
> I'm not sure if my ISP offers time servers, but Apple and MS both run time servers which are publicly accessible (presumably from any o/s).
> 
> I've never changed my laptop from its default, to sync with time.euro.apple.com, but my Linux boxes all use the public ntp pool, so I was surprised to read the other comments claiming the latter to be inaccurate.
> 
> Whenever I restart /etc/init.d/ntpd on my Linux boxes I can see their time match that of my laptop, as consistent as I can see, i.e. less than a second's difference between them.


ntpd has some wicked amazing optimizations built in, much more so if you
use multiple upstream sources. If one of them drifts, the software is
able to recognize it and defer instead to other sources that seem more
stable. It's like magic, the dodgy data tends to fall out of the system
leaving just the good data. Which is exactly what you want when using
volunteer resources of unknown and variable quality.

I'd compare the public ntp pool to a privateer race team - they can be
awesome, do amazing things with limited resources and often win races.
But for consistency and the best of the best, you need the Honda and
Yamaha factory teams (complete with obscene budgets).

For laptop, desktop and even most company's server needs, the public ntp
pool is perfectly good enough, which is what I think you observe in your
environment.


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 26/04/2013 20:54, Paul Hartman wrote:
> On Thu, Apr 25, 2013 at 9:33 AM, Nick Khamis <symack@gmail.com> wrote:
>> Hello Everyone,
>>
>> We are trying to sync our server's time with an accurate ntp
>> server, and was wondering which of the many solutions are
>> considered viable. I did see the
>> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
>> Our services are quite time sensitive.
> 
> I think the classic method is to use net-misc/ntp
> 
> See the extensive article at http://en.gentoo-wiki.com/wiki/NTP for
> great examples and description.
> 

Do none of us here ever deal with Windows? :-)

I notice that no-one has yet mentioned that Windows does not do ntp, as
Windows does not do time right, doesn't do timezones right and I
strongly suspect can't even do dates right (this latter still unproven)

Windows time servers need some magic Microsoft thing called ENTP which
is in no way related to the ntp we all know and love

-- 
Alan McKinnon
alan.mckinnon@gmail.com

[gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[the guard]

Пятница, 26 апреля 2013, 22:41 +02:00 от Alan McKinnon <alan.mckinnon@gmail.com>:
> On 26/04/2013 20:54, Paul Hartman wrote:
> > On Thu, Apr 25, 2013 at 9:33 AM, Nick Khamis <symack@gmail.com> wrote:
> >> Hello Everyone,
> >>
> >> We are trying to sync our server's time with an accurate ntp
> >> server, and was wondering which of the many solutions are
> >> considered viable. I did see the
> >> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
> >> Our services are quite time sensitive.
> > 
> > I think the classic method is to use net-misc/ntp
> > 
> > See the extensive article at http://en.gentoo-wiki.com/wiki/NTP for
> > great examples and description.
> > 
> 
> Do none of us here ever deal with Windows? :-)
> 
> I notice that no-one has yet mentioned that Windows does not do ntp, as
> Windows does not do time right, doesn't do timezones right and I
> strongly suspect can't even do dates right (this latter still unproven)
> 
> Windows time servers need some magic Microsoft thing called ENTP which
> is in no way related to the ntp we all know and love
> 
It refuses to adjust time if you have a wrong date.   timezone is set in your system

Re: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 26/04/2013 22:46, the guard wrote:
> 
> 
> 
> Пятница, 26 апреля 2013, 22:41 +02:00 от Alan McKinnon <alan.mckinnon@gmail.com>:
>> On 26/04/2013 20:54, Paul Hartman wrote:
>>> On Thu, Apr 25, 2013 at 9:33 AM, Nick Khamis <symack@gmail.com> wrote:
>>>> Hello Everyone,
>>>>
>>>> We are trying to sync our server's time with an accurate ntp
>>>> server, and was wondering which of the many solutions are
>>>> considered viable. I did see the
>>>> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
>>>> Our services are quite time sensitive.
>>>
>>> I think the classic method is to use net-misc/ntp
>>>
>>> See the extensive article at http://en.gentoo-wiki.com/wiki/NTP for
>>> great examples and description.
>>>
>>
>> Do none of us here ever deal with Windows? :-)
>>
>> I notice that no-one has yet mentioned that Windows does not do ntp, as
>> Windows does not do time right, doesn't do timezones right and I
>> strongly suspect can't even do dates right (this latter still unproven)
>>
>> Windows time servers need some magic Microsoft thing called ENTP which
>> is in no way related to the ntp we all know and love
>>
> It refuses to adjust time if you have a wrong date.   timezone is set in your system
> 


I was thinking more along the lines of how Windows has no concept of UTC
set in the hw clock and a local timezone, and how timezones are odd
things like Harare/Pretoria instead of the official names like
SAST GMT+2 as set by the scientific timekeeping community.

How about daylight savings? Can Windows deal with that? Other than by
just shoving the clock back and forward by an hour on the right days?


-- 
Alan McKinnon
alan.mckinnon@gmail.com

[gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[the guard]

Пятница, 26 апреля 2013, 22:54 +02:00 от Alan McKinnon <alan.mckinnon@gmail.com>:
> On 26/04/2013 22:46, the guard wrote:
> > 
> > 
> > 
> > Пятница, 26 апреля 2013, 22:41 +02:00 от Alan McKinnon <alan.mckinnon@gmail.com>:
> >> On 26/04/2013 20:54, Paul Hartman wrote:
> >>> On Thu, Apr 25, 2013 at 9:33 AM, Nick Khamis <symack@gmail.com> wrote:
> >>>> Hello Everyone,
> >>>>
> >>>> We are trying to sync our server's time with an accurate ntp
> >>>> server, and was wondering which of the many solutions are
> >>>> considered viable. I did see the
> >>>> http://en.gentoo-wiki.com/wiki/Time_Synchronization.
> >>>> Our services are quite time sensitive.
> >>>
> >>> I think the classic method is to use net-misc/ntp
> >>>
> >>> See the extensive article at http://en.gentoo-wiki.com/wiki/NTP for
> >>> great examples and description.
> >>>
> >>
> >> Do none of us here ever deal with Windows? :-)
> >>
> >> I notice that no-one has yet mentioned that Windows does not do ntp, as
> >> Windows does not do time right, doesn't do timezones right and I
> >> strongly suspect can't even do dates right (this latter still unproven)
> >>
> >> Windows time servers need some magic Microsoft thing called ENTP which
> >> is in no way related to the ntp we all know and love
> >>
> > It refuses to adjust time if you have a wrong date.   timezone is set in your system
> > 
> 
> 
> I was thinking more along the lines of how Windows has no concept of UTC
> set in the hw clock and a local timezone, and how timezones are odd
> things like Harare/Pretoria instead of the official names like
> SAST GMT+2 as set by the scientific timekeeping community.
> 
> How about daylight savings? Can Windows deal with that? Other than by
> just shoving the clock back and forward by an hour on the right days?
> 
> 
All I can say that XP didn't understand our polititians when they cancelled "summer time"
daylight saving. btw I saw a good quote in this list. something like  "only a white man can believe 
that by tearing a blanket at the bottom and attaching it on the top he will make tha blanket longer"????

Re: [gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 26/04/2013 23:02, the guard wrote:

>> I was thinking more along the lines of how Windows has no concept of UTC
>> set in the hw clock and a local timezone, and how timezones are odd
>> things like Harare/Pretoria instead of the official names like
>> SAST GMT+2 as set by the scientific timekeeping community.
>>
>> How about daylight savings? Can Windows deal with that? Other than by
>> just shoving the clock back and forward by an hour on the right days?
>>
>>
> All I can say that XP didn't understand our polititians when they cancelled "summer time"
> daylight saving. btw I saw a good quote in this list. something like  "only a white man can believe 
> that by tearing a blanket at the bottom and attaching it on the top he will make tha blanket longer"????
> 

XP didn't understand our politicians either, but we are a special case
amongst special cases. Nothing in this entire universe understands *our*
politicians, so XP gets a free pass on that one here :-)

And that's a funny joke, but not really accurate. Daylight savings is
designed to have the big orange ball visible in the sky for the maximum
amount of time whilst people are working at their daily 9 to 5. The day
doesn't get any longer, you just shift the darkness part forwards and
backwards.

I wasn't born here in Africa and didn't spend primary school years here
either. But I distinctly recall having to walk to school in the snow and
in the dark to geet their before 9 o'clock. Not fun. DST would have helped.


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Server system date synchronizaion

[Nick Khamis]

On 4/26/13, Alan McKinnon <alan.mckinnon@gmail.com> wrote:
> On 26/04/2013 19:11, Nick Khamis wrote:
>>>> >> Thank you so much for your response, and I totally understand the
>>>> >> effort vs. benefit challenge. However, is it really that much
>>>> >> trouble/unstable to setup our own ntp
>>>> >> server that syncs with our local isp, and have our internal network
>>>> >> sync
>>>> >> on it?
>>> >
>>> >
>>> > No, it's not THAT much effort. You can get by with installing ntpd on
>>> > a
>>> > single machine, pointing it at the upstream time server and pointing
>>> > all
>>> > your clients to it. It's clearly recorded in the config file, you
>>> > can't
>>> > go wrong.
>>> >
>>> > It's understanding how this weird thing called time works that is the
>>> > issue. Take for example leap seconds..... urggggggggggg...
>>> >
>>> > The basic question I suppose is why do you want to do it this way?
>>> > What
>>> > do you feel you will gain by doing it yourself?
>>> >
>>> >
>>> > --
>>> > Alan McKinnon
>>> > alan.mckinnon@gmail.com
>>> >
>>> >
>>> >
>> Hello Alan,
>>
>> Thank you so much for your time. Our voip cluster time always vary for
>> some reason....
>> And with long distance, that could mean upwards to a dollar a call.
>
>
> Ah, OK. That changes things quite a bit. I have a little bit of
> experience with that - I work for a large ISP, we have a large VOIP
> department and we run a stratum 2 time server that serves most of the
> country.
>
> First things first: you can't just stick any old upstream ntp server in
> your config and walk away. You are then reliant on the quality of that
> upstream, and far too often other time servers operate on a "good
> enough" policy - if it's accurate to about a second, it's good enough
> (and for desktop users i.e. most ISP clients, it is good enough).
>
> I don't know how big your operation is, if you have budget I suggest you
> invest in a proper master time source that is GPS-driven. We have a
> Symmetricom (http://www.symmetricom.com) but it's a mature market with
> several vendors. Shop around, prices are less than you'd expect (about
> the same as a decent mid-range server and much less than Cisco's
> routers...)
>
> Weather can get in the way, so back up the device with a decent second
> upstream. I have a good one available run by the Science and Technology
> Research part of the Dept of Trade and Industry and the third option is
> all the other big ISPs around.
>
> Depending on your accuracy needs you could get away without the GPS unit
> and just use a good upstream, but I'd fight for the budget for it - tell
> management it puts control of billing back in your hands, they always
> fall for that one :-)
>
> So the summary would be that I reckon ntpd will do what you want as long
> as you chose good reliable time sources. With that in hand, the config
> is easy as rather well documented. Shout here ont he list if you need a
> hand with this when you come to deployment time
>
>
>
>
> --
> Alan McKinnon
> alan.mckinnon@gmail.com
>
>
>

Any suggestions for a "reliable", use that word cautiously ntp server.
Requests are coming from canada. Was there not a project that dealt
with setting up a network across the globe just for serving up NTP
services? Did that marvelous idea die out?

N.

Re: [gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 591 bytes --]

On Fri, 26 Apr 2013 23:10:46 +0200, Alan McKinnon wrote:

> I wasn't born here in Africa and didn't spend primary school years here
> either. But I distinctly recall having to walk to school in the snow and
> in the dark to geet their before 9 o'clock. Not fun. DST would have
> helped.

No it wouldn't - DST makes it darker in the morning. When I was about 11,
the government experimented with using BST all year round. One of the
reasons given for not doing it was that kids would have to go to school
in the dark.


-- 
Neil Bothwick

Top Oxymorons Number 22: Childproof

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Paul Hartman]

On Fri, Apr 26, 2013 at 3:54 PM, Alan McKinnon <alan.mckinnon@gmail.com> wrote:
> On 26/04/2013 22:46, the guard wrote:
>> Пятница, 26 апреля 2013, 22:41 +02:00 от Alan McKinnon <alan.mckinnon@gmail.com>:
>>> Do none of us here ever deal with Windows? :-)
>>>
>>> I notice that no-one has yet mentioned that Windows does not do ntp, as
>>> Windows does not do time right, doesn't do timezones right and I
>>> strongly suspect can't even do dates right (this latter still unproven)
>>>
>>> Windows time servers need some magic Microsoft thing called ENTP which
>>> is in no way related to the ntp we all know and love
>>>
>> It refuses to adjust time if you have a wrong date.   timezone is set in your system
>>
>
>
> I was thinking more along the lines of how Windows has no concept of UTC
> set in the hw clock and a local timezone, and how timezones are odd
> things like Harare/Pretoria instead of the official names like
> SAST GMT+2 as set by the scientific timekeeping community.
>
> How about daylight savings? Can Windows deal with that? Other than by
> just shoving the clock back and forward by an hour on the right days?

I've used windows for the past 25,000+ work hours at my job (I wish
that were an exaggeration) in an all-Microsoft corporate environment.
I dare not declare myself an expert in anything Windows so as not to
encourage more of it. :)

AFAIK the windows time service (w32time) does everything internally
and between machines using UTC, but translates to/from local time for
updating the hardware clock and the OS time. When daylight saving time
happens it just changes the clock, though I have heard of some sites
where the time change does not occur until the next time sync happens.
If DST happens when the machine is powered off, it changes it at the
next reboot (and usually pops up a little window to let you know what
has happened). Sometimes if you reboot multiple times on a DST
changeover day it can adjust the clock repeatedly...

If you haven't installed Windows Updates or are using an unsupported
version, your DST and time zone info may be outdated. For example, in
the US about 10 years ago they changed the start and end of DST by a
few weeks. Any devices using the old logic will be wrong for about a
month out of the year. If someone manually "fixes" the time on their
workstation, it will be correct until it changes itself and then it'll
be wrong again. :)

Also, being Windows, people tend to set the wrong time zone, don't
check the "use daylight saving" box, choose Central America
(continent) instead of Central US (country) time zone, etc. Then they
send out meeting invitations in Outlook and the time gets shifted by
the Exchange server and everybody shows up to a conference room an
hour early, except for the person who organized the meeting,
naturally.

Time sync has been built into Windows since Win 2000, and machines who
are part of a domain sync time with their domain controller using some
proprietary protocol called NT5DS. If you have admin rights you can
edit the registry and change it to use plain old NTP and sync with a
regular NTP server. The DC can sync with other DCs or standard NTP
server(s) over the internet. Home machines w/o a domain can set an NTP
server in the date and time settings without messing with the
registry, I think. (I don't use Windows at home.)

The time sync service by default changes the time gradually, taking up
to an hour to make the adjustment when there is a difference. Not sure
if there is an upper limit where it refuses to adjust if it's "too
wrong". You can also force an immediate sync in those cases.

There is a multi-purpose time utility built-in to windows called
w32tm.exe that lets you do various time operations, giving some
insight into the way Windows sees the world. I can do things like:

C:\Windows\system32>w32tm /tz
Time zone: Current:TIME_ZONE_ID_DAYLIGHT Bias: 360min (UTC=LocalTime+Bias)
  [Standard Name:"Central Standard Time" Bias:0min Date:(M:11 D:1 DoW:0)]
  [Daylight Name:"Central Daylight Time" Bias:-60min Date:(M:3 D:2 DoW:0)]

The interesting part there is UTC=LocalTime+Bias. So that seems to be
how they handle that. The other lines show what it knows about when
DST kicks in and the additional bias.

C:\Windows\system32>w32tm /query /status
Leap Indicator: 0(no warning)
Stratum: 4 (secondary reference - syncd by (S)NTP)
Precision: -6 (15.625ms per tick)
Root Delay: 0.2329102s
Root Dispersion: 0.3298777s
ReferenceId: 0x0A010046 (source IP:  10.1.0.70)
Last Successful Sync Time: 4/26/2013 10:37:44 AM
Source: DC1.example.com
Poll Interval: 15 (32768s)

Tells me about the time sync status on my workstation and info about
the last sync.

C:\Windows\system32>w32tm /stripchart /computer:time-a.nist.gov /samples:10
Tracking time-a.nist.gov [129.6.15.28:123].
Collecting 10 samples.
The current time is 4/26/2013 4:08:03 PM.
16:08:03 d:+00.0467925s o:-00.2902514s  [                          *|
                         ]
16:08:05 d:+00.0623842s o:-00.2958840s  [                          *|
                         ]
16:08:07 d:+00.0311857s o:-00.2882881s  [                          *|
                         ]
16:08:09 d:+00.0467864s o:-00.2917686s  [                          *|
                         ]
16:08:11 d:+00.0467916s o:-00.2846576s  [                          *|
                         ]
16:08:13 d:+00.0467925s o:-00.2963050s  [                          *|
                         ]
16:08:15 d:+00.0467925s o:-00.2963291s  [                          *|
                         ]
16:08:17 d:+00.0467919s o:-00.2949080s  [                          *|
                         ]
16:08:19 d:+00.0467897s o:-00.2966004s  [                          *|
                         ]
16:08:21 d:+00.0311927s o:-00.2887318s  [                          *|
                         ]

Measuring differences between my machine and another one. Notice here
I'm querying one of the same standard time servers I sync with on my
Gentoo machines.

Now we return to our regularly-scheduled programming...

Re: [gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Peter Humphrey]

On Friday 26 April 2013 22:43:10 Neil Bothwick wrote:
> On Fri, 26 Apr 2013 23:10:46 +0200, Alan McKinnon wrote:
> > I wasn't born here in Africa and didn't spend primary school years here
> > either. But I distinctly recall having to walk to school in the snow
> > and in the dark to geet their before 9 o'clock. Not fun. DST would
> > have helped.
> 
> No it wouldn't - DST makes it darker in the morning. When I was about 11,
> the government experimented with using BST all year round. One of the
> reasons given for not doing it was that kids would have to go to school
> in the dark.

... and the children's accident rate in Scotland shot up.

And what is this idea of saving daylight? Only an American could conceive of 
such a nonsense (I hope).

-- 
Peter

Re: [gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[William Kenworthy]

On 27/04/13 09:20, Peter Humphrey wrote:
> On Friday 26 April 2013 22:43:10 Neil Bothwick wrote:
>> On Fri, 26 Apr 2013 23:10:46 +0200, Alan McKinnon wrote:
>>> I wasn't born here in Africa and didn't spend primary school years here
>>> either. But I distinctly recall having to walk to school in the snow
>>> and in the dark to geet their before 9 o'clock. Not fun. DST would
>>> have helped.
>>
>> No it wouldn't - DST makes it darker in the morning. When I was about 11,
>> the government experimented with using BST all year round. One of the
>> reasons given for not doing it was that kids would have to go to school
>> in the dark.
> 
> ... and the children's accident rate in Scotland shot up.
> 
> And what is this idea of saving daylight? Only an American could conceive of 
> such a nonsense (I hope).
> 

I wish it were so ... every now and again they decide to try it -
usually because one lot of pollies want to get the drop on the other
and/or distract the sheeple with an "issue"

zdump -v Australia/Perth

The pollies (bless their little black hearts) decided to implement a
trial with only a few weeks notice! - Linux/Unix had the updates within
a day of the specs, distros followed with formal a couple of weeks
later.  MS took 12 months and exchange calendars where I work corrupted
and had the be manually reentered (and then defaulted to the Ulan Bator
timezone in Mongolia as they couldnt get windows to do it locally - yes
they have a MS support contract and its a mainly MS shop).  Same going
back ... dont know what it cost in lost productivity, mistakes and other
problems but it wasn't small.

After the three year trial the pollies went to a referendum and said
"ok, you have had 3 years and don't you like it now your used to it?
..." and it was thrown out yet again :)

Even once the fixes were in windows, each change point was problematic
for some/many in IT.

BillK

Re[2]: [gentoo-user] Server system date synchronizaion

[Andrew Lowe]

On 04/27/13 09:20, Peter Humphrey wrote:
> On Friday 26 April 2013 22:43:10 Neil Bothwick wrote:
>> On Fri, 26 Apr 2013 23:10:46 +0200, Alan McKinnon wrote:
>>> I wasn't born here in Africa and didn't spend primary school years here
>>> either. But I distinctly recall having to walk to school in the snow
>>> and in the dark to geet their before 9 o'clock. Not fun. DST would
>>> have helped.
>>
>> No it wouldn't - DST makes it darker in the morning. When I was about 11,
>> the government experimented with using BST all year round. One of the
>> reasons given for not doing it was that kids would have to go to school
>> in the dark.
>
> ... and the children's accident rate in Scotland shot up.
>
> And what is this idea of saving daylight? Only an American could conceive of
> such a nonsense (I hope).
>
	And the curtains will fade quicker with all that extra saved sunlight. 
And has anyone thought of the dairy cows? Farmers get up at 4am to milk 
them, but the cows will get confused because they don't have watches so 
when the farmer turns up at 4am, DST, the cows won't be there........ 
and on it goes, drivel from the great unwashed masses who can't wrap 
their head around the concept of shifting the clock 1 hour.

	Get over it and enjoy the extra hour in the evening. But then again I'm 
in Australia where we have things to enjoy as distinct from those who 
appear to be in the UK and complaining :) And lets be realistic about 
childrens accident rates due to daylight savings, hasn't anyone heard of 
Darwinian theory....

	Regards,
		Andrew

Re: [gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1455 bytes --]

On Sat, 27 Apr 2013 02:20:08 +0100, Peter Humphrey wrote:

> > No it wouldn't - DST makes it darker in the morning. When I was about
> > 11, the government experimented with using BST all year round. One of
> > the reasons given for not doing it was that kids would have to go to
> > school in the dark.  
> 
> ... and the children's accident rate in Scotland shot up.

They have re-examined the data and found that that was not necessarily
true. While the accident rate in the morning did go up, the afternoon
rate went down by more, because drivers are more alert in the morning so
cope with the darkness better.

It wasn't only Scotland, in fact they aren't affected that much anyway. I
worked in Dundee one December and it was dark until almost 10am, without
DST. When they trialled winter DST, I was going to school in the dark in
London.
 
> And what is this idea of saving daylight? Only an American could
> conceive of such a nonsense (I hope).

It was the Germans, during WW1, quickly copied by Britain. The idea was
to improve the productivity of the factories. And you are saving
daylight, you are saving up an hour of daylight that you would otherwise
sleep through and spending it at a more suitable time a few hours later.

No one suggested a Daylight Bank :-O


-- 
Neil Bothwick

Mouse: (n.) an input device used by management to force computer users to
       keep at least a part of their desks clean.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 26/04/2013 23:28, Nick Khamis wrote:
> On 4/26/13, Alan McKinnon <alan.mckinnon@gmail.com> wrote:
>> On 26/04/2013 19:11, Nick Khamis wrote:
>>>>>>> Thank you so much for your response, and I totally understand the
>>>>>>> effort vs. benefit challenge. However, is it really that much
>>>>>>> trouble/unstable to setup our own ntp
>>>>>>> server that syncs with our local isp, and have our internal network
>>>>>>> sync
>>>>>>> on it?
>>>>>
>>>>>
>>>>> No, it's not THAT much effort. You can get by with installing ntpd on
>>>>> a
>>>>> single machine, pointing it at the upstream time server and pointing
>>>>> all
>>>>> your clients to it. It's clearly recorded in the config file, you
>>>>> can't
>>>>> go wrong.
>>>>>
>>>>> It's understanding how this weird thing called time works that is the
>>>>> issue. Take for example leap seconds..... urggggggggggg...
>>>>>
>>>>> The basic question I suppose is why do you want to do it this way?
>>>>> What
>>>>> do you feel you will gain by doing it yourself?
>>>>>
>>>>>
>>>>> --
>>>>> Alan McKinnon
>>>>> alan.mckinnon@gmail.com
>>>>>
>>>>>
>>>>>
>>> Hello Alan,
>>>
>>> Thank you so much for your time. Our voip cluster time always vary for
>>> some reason....
>>> And with long distance, that could mean upwards to a dollar a call.
>>
>>
>> Ah, OK. That changes things quite a bit. I have a little bit of
>> experience with that - I work for a large ISP, we have a large VOIP
>> department and we run a stratum 2 time server that serves most of the
>> country.
>>
>> First things first: you can't just stick any old upstream ntp server in
>> your config and walk away. You are then reliant on the quality of that
>> upstream, and far too often other time servers operate on a "good
>> enough" policy - if it's accurate to about a second, it's good enough
>> (and for desktop users i.e. most ISP clients, it is good enough).
>>
>> I don't know how big your operation is, if you have budget I suggest you
>> invest in a proper master time source that is GPS-driven. We have a
>> Symmetricom (http://www.symmetricom.com) but it's a mature market with
>> several vendors. Shop around, prices are less than you'd expect (about
>> the same as a decent mid-range server and much less than Cisco's
>> routers...)
>>
>> Weather can get in the way, so back up the device with a decent second
>> upstream. I have a good one available run by the Science and Technology
>> Research part of the Dept of Trade and Industry and the third option is
>> all the other big ISPs around.
>>
>> Depending on your accuracy needs you could get away without the GPS unit
>> and just use a good upstream, but I'd fight for the budget for it - tell
>> management it puts control of billing back in your hands, they always
>> fall for that one :-)
>>
>> So the summary would be that I reckon ntpd will do what you want as long
>> as you chose good reliable time sources. With that in hand, the config
>> is easy as rather well documented. Shout here ont he list if you need a
>> hand with this when you come to deployment time
>>
>>
>>
>>
>> --
>> Alan McKinnon
>> alan.mckinnon@gmail.com
>>
>>
>>
> 
> Any suggestions for a "reliable", use that word cautiously ntp server.
> Requests are coming from canada. Was there not a project that dealt
> with setting up a network across the globe just for serving up NTP
> services? Did that marvelous idea die out?

Isn't that what pool.ntp.org does?

As for reliable, I'm not familiar with how Canada has set itself up, but
most Western governments have a "Science and Technology" department or
NGO and most run time servers to serve the local scientific community.
They might not let you sync to their server (stratum 1 providers are
touchy) but someone will sync to it, and they in turn may provide a free
time service.

Start by Googling "stratum 1 time server Canada" and see where that
takes you. Really, this stuff isn't hard and you will be up and running
in no time. The hard part is when *you* provide a public service and
need to pay attention to the insane amount of detail inherent in this
subject.


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 27/04/2013 03:20, Peter Humphrey wrote:
> On Friday 26 April 2013 22:43:10 Neil Bothwick wrote:
>> On Fri, 26 Apr 2013 23:10:46 +0200, Alan McKinnon wrote:
>>> I wasn't born here in Africa and didn't spend primary school years here
>>> either. But I distinctly recall having to walk to school in the snow
>>> and in the dark to geet their before 9 o'clock. Not fun. DST would
>>> have helped.
>>
>> No it wouldn't - DST makes it darker in the morning. When I was about 11,
>> the government experimented with using BST all year round. One of the
>> reasons given for not doing it was that kids would have to go to school
>> in the dark.
> 
> ... and the children's accident rate in Scotland shot up.
> 
> And what is this idea of saving daylight? Only an American could conceive of 
> such a nonsense (I hope).
> 

It's not saving daylight, taken literally it's a misnomer.

"Daylight Savings" is just easier to say and get an acronym for than
"Having more of the hours in a day when you are up awake and trying to
do work happen when it's light rather than dark"

-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 27/04/2013 05:44, Andrew Lowe wrote:
> Get over it and enjoy the extra hour in the evening. But then again I'm
> in Australia where

[snip here]

OK, stop right there. I see where the disconnect comes in.

You are in Australia. The sun happens to shine in Australia. It shines a
lot there.
I am in South Africa. The sun happens to shine a lot in South Africa. It
shines a lot here.

Neil is in England. The sun never shines in England. It makes the
English confused and fries their brains.



-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 27/04/2013 00:11, Paul Hartman wrote:
> On Fri, Apr 26, 2013 at 3:54 PM, Alan McKinnon <alan.mckinnon@gmail.com> wrote:
>> On 26/04/2013 22:46, the guard wrote:
>>> Пятница, 26 апреля 2013, 22:41 +02:00 от Alan McKinnon <alan.mckinnon@gmail.com>:
>>>> Do none of us here ever deal with Windows? :-)
>>>>
>>>> I notice that no-one has yet mentioned that Windows does not do ntp, as
>>>> Windows does not do time right, doesn't do timezones right and I
>>>> strongly suspect can't even do dates right (this latter still unproven)
>>>>
>>>> Windows time servers need some magic Microsoft thing called ENTP which
>>>> is in no way related to the ntp we all know and love
>>>>
>>> It refuses to adjust time if you have a wrong date.   timezone is set in your system
>>>
>>
>>
>> I was thinking more along the lines of how Windows has no concept of UTC
>> set in the hw clock and a local timezone, and how timezones are odd
>> things like Harare/Pretoria instead of the official names like
>> SAST GMT+2 as set by the scientific timekeeping community.
>>
>> How about daylight savings? Can Windows deal with that? Other than by
>> just shoving the clock back and forward by an hour on the right days?
> 
> I've used windows for the past 25,000+ work hours at my job (I wish
> that were an exaggeration) in an all-Microsoft corporate environment.
> I dare not declare myself an expert in anything Windows so as not to
> encourage more of it. :)

[snip much detail about Windows tying itself in knots to do something
quite simple]

> Now we return to our regularly-scheduled programming...

In comparison I have it easy :-)

The last time there was a leap-second all I had to do was check our time
servers tracked upstream wrt leap-*, and checked that my team's machines
did the same. ClusterSSH, bash, grep, sed and awk made this an exercise
in on-liner skills :-)

I then told the reat of the company using Unix to do the same, and the
whole thing was a non-event.

Now for the Windows fellows and the idiot running the Domain Controllers
in OurAmazingParentCompanyWhoThinkTheyAreCool(tm). Apparently they had a
torrid time of it, especially as they ignored all heads-up
communications from us. I don't know how they managed to fix all the
Windows workstations as we were quite happy to return that favour.


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Tanstaafl]

On 2013-04-26 5:10 PM, Alan McKinnon <alan.mckinnon@gmail.com> wrote:
> I wasn't born here in Africa and didn't spend primary school years here
> either. But I distinctly recall having to walk to school in the snow and
> in the dark to geet their before 9 o'clock. Not fun. DST would have helped.

But what would make more sense (at least in my mind) would be to have 
official 'working hours' changes, and adjust those, rather than change 
the clocks - ie, in your example, instead of school starting at 9:00am, 
it switches to start at 10:00am.

Re: [gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 27/04/2013 18:24, Tanstaafl wrote:
> On 2013-04-26 5:10 PM, Alan McKinnon <alan.mckinnon@gmail.com> wrote:
>> I wasn't born here in Africa and didn't spend primary school years here
>> either. But I distinctly recall having to walk to school in the snow and
>> in the dark to geet their before 9 o'clock. Not fun. DST would have
>> helped.
> 
> But what would make more sense (at least in my mind) would be to have
> official 'working hours' changes, and adjust those, rather than change
> the clocks - ie, in your example, instead of school starting at 9:00am,
> it switches to start at 10:00am.
> 

That's probably a better way overall.

It would suit the dairy cows!


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Server system date synchronizaion

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 714 bytes --]

On Sat, 27 Apr 2013 17:15:37 +0200, Alan McKinnon wrote:

> You are in Australia. The sun happens to shine in Australia. It shines a
> lot there.
> I am in South Africa. The sun happens to shine a lot in South Africa. It
> shines a lot here.
> 
> Neil is in England. The sun never shines in England. It makes the
> English confused and fries their brains.

Of course it shines over here, it does it non-stop every time I go abroad
on holiday :(

For the record, I wasn't one of those complaining about DST, it makes
little difference to me whether or not there is a sun the other side of
those rain clouds.


-- 
Neil Bothwick

Copper wire was invented by two Scotsmen fighting over a penny!

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]

Re: [gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Alan McKinnon]

On 26/04/2013 23:43, Neil Bothwick wrote:
> On Fri, 26 Apr 2013 23:10:46 +0200, Alan McKinnon wrote:
> 
>> I wasn't born here in Africa and didn't spend primary school years here
>> either. But I distinctly recall having to walk to school in the snow and
>> in the dark to geet their before 9 o'clock. Not fun. DST would have
>> helped.
> 
> No it wouldn't - DST makes it darker in the morning. When I was about 11,
> the government experimented with using BST all year round. One of the
> reasons given for not doing it was that kids would have to go to school
> in the dark.
> 
> 


Hmmm. DST as punted here was couched in reverse terms - have kids go to
school when it was light, as it would still be light in the early
afternoon when they went home.

DST never took off here, it would only be useful in the Cape down south
and only for about a month or so of the year.


-- 
Alan McKinnon
alan.mckinnon@gmail.com

Re: [gentoo-user] Re[2]: [gentoo-user] Re[2]: [gentoo-user] Server system date synchronizaion

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 764 bytes --]

On Sat, 27 Apr 2013 12:24:53 -0400, Tanstaafl wrote:

> But what would make more sense (at least in my mind) would be to have 
> official 'working hours' changes, and adjust those, rather than change 
> the clocks - ie, in your example, instead of school starting at 9:00am, 
> it switches to start at 10:00am.

That would mean lots of separate changes, including things that are
regulated by time, such as licencing hours (which, incidentally, were
introduced here in the same Act of Parliament as DST, and for the same
reason).

A single, well documented, change to clocks is a lot simpler than every
business and organisation having date-dependent trading hours.


-- 
Neil Bothwick

Anything worth fighting for is worth fighting dirty for.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]