Re: [smartos-discuss] Power Management on Modern CPUs

[Paul B. Henson]

On Thu, Oct 06, 2016 at 03:42:36AM +0100, Adam Richmond-Gordon wrote:

> Thank you for pointing that out - I had it in my head that a single
> log device would be safe.

There's only one failure mode (AFAIK) where a single log device will
cause data loss; if your box crashes or has an unclean shutdown (say due
to a power failure) while there are uncommitted entries on the log
device, and the device fails before the system comes back online to
process them.

If the device dies while the system is running the log will fall back to
being on the pool and any uncommitted entries are still in memory and
won't be lost. If the device dies after a clean shutdown or poweroff
there won't be any uncommitted entries on it and when the system comes
up it will fail the device and again just fall back to an on-pool log.

So while it's true that a single log device is non-redundant, it's a lot
less "not safe" than say a non-redundant pool. You'd have to be pretty
unlucky to actually have data loss from losing a non-redundant log
device. Of course, depending on the importance of your data, that might
not be a risk you want to take. But for a budget sensitive system, a
single high-cost SSD for a log isn't an insane configuration if you
think the odds of your system crashing/powering off dirty at the exact
same time your log device dies are pretty low.


---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Re: [smartos-discuss] Power Management on Modern CPUs

[Adam Richmond-Gordon]

Hi Daniel,

> Just worth mentioning in case you weren't aware of the consequences.
> 
Thank you for pointing that out - I had it in my head that a single log device 
would be safe. I have now rectified this;
# zpool status
  pool: zones
 state: ONLINE
  scan: scrub repaired 0 in 4h16m with 0 errors on Sun Sep 11 17:36:10 2016
config:

NAME   STATE READ WRITE CKSUM
zones  ONLINE   0 0 0
  raidz1-0 ONLINE   0 0 0
c1t5000CCA07D010981d0  ONLINE   0 0 0
c1t5000CCA07D0109B9d0  ONLINE   0 0 0
c1t5000CCA07D010A75d0  ONLINE   0 0 0
c1t5000CCA07D01270Dd0  ONLINE   0 0 0
  raidz1-1 ONLINE   0 0 0
c1t5000CCA07D0109A9d0  ONLINE   0 0 0
c1t5000CCA07D012DE9d0  ONLINE   0 0 0
c1t5000CCA07D012EE1d0  ONLINE   0 0 0
c1t5000CCA07D013131d0  ONLINE   0 0 0
  raidz1-3 ONLINE   0 0 0
c0t5000C500971AAED3d0  ONLINE   0 0 0
c0t5000C500971AD5CBd0  ONLINE   0 0 0
c0t5000C500971AF8ABd0  ONLINE   0 0 0
c0t5000C500971BDEFBd0  ONLINE   0 0 0
  raidz1-4 ONLINE   0 0 0
c0t5000C500971BE8DFd0  ONLINE   0 0 0
c0t5000C500971BE267d0  ONLINE   0 0 0
c0t5000C50097190C4Fd0  ONLINE   0 0 0
c0t5000C5009719301Bd0  ONLINE   0 0 0
logs
  mirror-2 ONLINE   0 0 0
c2t4d0 ONLINE   0 0 0
c2t5d0 ONLINE   0 0 0

errors: No known data errors

Adam

> On 6 Oct 2016, at 00:20, Daniel Carosone  wrote:
> 
> Regardless of anything else, you have a non-redundant log device for your 
> pool, and thus a non-redundant pool at least with respect to recent committed 
> transactions.
> 
> Just worth mentioning in case you weren't aware of the consequences.  
> Whatever the root cause of the present issue, you will need to rethink the 
> ssd usage.
> 
> smartos-discuss | Archives 
>   
>  | 
> Modify  Your Subscription   
> 



---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Re: [smartos-discuss] Windows KVM with many cores showing high load average

[Ian Collins]

On 10/ 6/16 01:35 PM, Ian Collins wrote:

On 10/ 6/16 12:24 PM, Matthew Parsons wrote:
Oh wow, I meant what version of Windows guests! (As you might expect, 
guest behavior can be rather different between, say, 2003 vs 2012 R2.)


It's 2012 R2.

Quick google shows that version probably came from 
https://download.joyent.com/pub/vmtools/,  so hopefully you're runing 
2012 :) - It's been a fair few months since I've had to deal w/ 
Windows under KVM, and it sounds like I don't know anything you don't 
- but just in case, I'd start at 
https://fedoraproject.org/wiki/Windows_Virtio_Drivers and try 
updating w/ the latest there.


I've just migrated the machine back to a single core test host and it 
is now using 39% of the host CPU, so that blows the lots of cores 
hypothesis.  Another 2012R2 guest on the same host is idling at 4%, so 
there's something going on the the VM.


Yep, I should have known: it was a (spit) java service...

--
Ian.




---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Re: [smartos-discuss] Windows KVM with many cores showing high load average

[Matthew Parsons]

Oh wow, I meant what version of Windows guests! (As you might expect, guest
behavior can be rather different between, say, 2003 vs 2012 R2.)

Quick google shows that version probably came from
https://download.joyent.com/pub/vmtools/,  so hopefully you're runing 2012
:) - It's been a fair few months since I've had to deal w/ Windows under
KVM, and it sounds like I don't know anything you don't - but just in case,
I'd start at https://fedoraproject.org/wiki/Windows_Virtio_Drivers and try
updating w/ the latest there.

(http://www.linux-kvm.org/page/WindowsGuestDrivers/Download_Drivers - for
more info)

On Wed, Oct 5, 2016 at 3:48 PM, Ian Collins  wrote:

> On 10/ 6/16 10:03 AM, Matthew Parsons wrote:
>
>> Have you installed the VirtIO drivers in windows? (And what version?) For
>> testing I'd try disabling/removing the guest NICs and just see if
>> interrupts die down.
>>
>
> Which version is often an issue!  There are so many out there and some
> work with one version of windows and not another!  The ISO I use is named
> "me-ws2012std-20130712.iso" and I've been using it for a couple of years so
> its origins are lost in the mists of time.
>
> Also (again for testing) perhaps reduce cores to the amount on a physical
>> CPU socket and assign/restrict to avoid crossing NUMA boundries.
>>
>
> The problem only becomes an issue when the core number gets high, as I
> said in my original post the load average almost quadruples when going from
> 16 to 32 cores.
>
> (I trust that whatever workload you're running benefits from that many
>> cores, but typically I'd keep 2 or so for the hypervisor/management/other.)
>>
> 
> The workload is compiling a large C and C++ code base, so the more cores
> the better.
> 
> Experimenting on a smaller machine shows the build times to cores ratios
> reflect those on bare metal, that is if I give the VM the full system
> picture (using qemu_extra_opts) build times are about 25% faster than
> giving it the number of physical cores (using vcpus).  For example to get
> optimum performance on a single quad core, use "vcpus": 1,
> "qemu_extra_opts": "-smp cpus=1,cores=4,threads=2".
> 
> Cheers,
> 
> --
> Ian.
> 



---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Re: [smartos-discuss] Power Management on Modern CPUs

[Daniel Carosone]

Regardless of anything else, you have a non-redundant log device for your
pool, and thus a non-redundant pool at least with respect to recent
committed transactions.

Just worth mentioning in case you weren't aware of the consequences.
Whatever the root cause of the present issue, you will need to rethink the
ssd usage.



---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Re: [smartos-discuss] Windows KVM with many cores showing high load average

[Matthew Parsons]

Have you installed the VirtIO drivers in windows? (And what version?) For
testing I'd try disabling/removing the guest NICs and just see if
interrupts die down. Also (again for testing) perhaps reduce cores to the
amount on a physical CPU socket and assign/restrict to avoid crossing NUMA
boundries.

(I trust that whatever workload you're running benefits from that many
cores, but typically I'd keep 2 or so for the hypervisor/management/other.)

On Tue, Oct 4, 2016 at 3:19 PM, Ian Collins  wrote:

> On 5/10/16 9:57 am, Tiraen wrote:
>
>> It all very much depends on the processor itself (and even the same CPU
>> may work in different).
>>
>> A little is not correct to compare in this regard, Linux systems and
>> windows
>>
>> Since they have fundamentally different ways of working with CPU (CLR in
>> windows and things like that)
>>
>> If in general - what you see - it's normal
>>
>>
> The load appears to come in the form of interrupts.  The Windows KVM zones
> are generating 32K interrupts/second while the Linux KVM generates 25...
> 
> --
> Ian.
> 



---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Re: [smartos-discuss] inet6 default gateway in non-global zones

[Cody Mello]

Hey Juraj,

On Wed, Oct 5, 2016 at 8:58 AM, Juraj Lutter  wrote:
> Yes, I know of -p but I thought that there
> is also another, SmartOS way.

There is the "gateways" field for NICs, which currently only supports
IPv4 addresses, but will eventually support IPv6. For now, the "route
-p add" method is probably the best thing to do. (Or set up some
script to run it at every reboot for you, which will be necessary for
lx zones.)

- Cody


---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Re: [smartos-discuss] inet6 default gateway in non-global zones

[Jorge Schrauwen]

This doesn't work in an lx zone however, unless you add some glue to 
reload the rules at boot.


---
~ sjorge

On 2016-10-05 17:54, Dan McDonald wrote:
On Oct 5, 2016, at 11:06 AM, Juraj Lutter  
wrote:


Hi,

just a quick question: Can one specify inet6 default route for 
joyent-branded zones? From what I understand, net-physical relies on 
NDP (which might not always be available.)


Is there any other systematic way?


The route(1M) command works for IPv6, and does have its "-p" option
(persistent storage).  Since native zones have actual backing store on
/etc/inet, it's a perfectly acceptable way of adding a persistent IPv6
route if the vmadm-style configuration doesn't do it.  For example:

route -p add -inet6 default fe80::dead:beef:feed:face

Hope this helps,
Dan




---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Re: [smartos-discuss] inet6 default gateway in non-global zones

[Juraj Lutter]

--
Juraj Lutter

> On 5 Oct 2016, at 17:54, Dan McDonald  wrote:
> 
> 
>> On Oct 5, 2016, at 11:06 AM, Juraj Lutter  wrote:
>> 
>> Hi,
>> 
>> just a quick question: Can one specify inet6 default route for 
>> joyent-branded zones? From what I understand, net-physical relies on NDP 
>> (which might not always be available.)
>> 
>> Is there any other systematic way?
> 
> The route(1M) command works for IPv6, and does have its "-p" option 
> (persistent storage).  Since native zones have actual backing store on 
> /etc/inet, it's a perfectly acceptable way of adding a persistent IPv6 route 
> if the vmadm-style configuration doesn't do it.  For example:
> 
>route -p add -inet6 default fe80::dead:beef:feed:face

Yes, I know of -p but I thought that there
is also another, SmartOS way. 


> 
> Hope this helps,
> Dan
> 


---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

Re: [smartos-discuss] inet6 default gateway in non-global zones

[Dan McDonald]

> On Oct 5, 2016, at 11:06 AM, Juraj Lutter  wrote:
> 
> Hi,
> 
> just a quick question: Can one specify inet6 default route for joyent-branded 
> zones? From what I understand, net-physical relies on NDP (which might not 
> always be available.)
> 
> Is there any other systematic way?

The route(1M) command works for IPv6, and does have its "-p" option (persistent 
storage).  Since native zones have actual backing store on /etc/inet, it's a 
perfectly acceptable way of adding a persistent IPv6 route if the vmadm-style 
configuration doesn't do it.  For example:

route -p add -inet6 default fe80::dead:beef:feed:face

Hope this helps,
Dan



---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com

[smartos-discuss] inet6 default gateway in non-global zones

[Juraj Lutter]

Hi,

just a quick question: Can one specify inet6 default route for joyent-branded 
zones? From what I understand, net-physical relies on NDP (which might not 
always be available.)

Is there any other systematic way?

Thanks.


--
  Juraj Lutter
Systems Engineer
+421 907 986 576

https://sk.linkedin.com/in/jurajlutter 

Erigones, s. r. o.

Company HQ: Priemyselná 4, 949 01 Nitra, Slovak Republic

Technical Department: Sliačska 1/D, 831 02 Bratislava, Slovak Republic

http://www.erigones.com/ 







---
smartos-discuss
Archives: https://www.listbox.com/member/archive/184463/=now
RSS Feed: https://www.listbox.com/member/archive/rss/184463/25769125-55cfbc00
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=25769125_secret=25769125-7688e9fb
Powered by Listbox: http://www.listbox.com