On 2015-01-27 16:46, Miuku wrote:
>
> VS_D_Media;2691587 Wrote:
>> Thanks a lot mate! We dont have lots of money but if I get to manage
>> that I will have to ask my boss to send you some money on paypal
>> hehehehe
> Thanks for the offer but I do this for free. If you guys want to spend
> money, send some to your local charity
Same here, I do this largely for fun.
And openSUSE as an organization doesnāt have the legal status to accept
donations, I understand.
Instead what is asked of people is to try contribute in many other ways,
like helping on forums or mail lists, or working on projects, like
packaging programs, translating/creating/maintaining programs,
documentation, web pagesā¦ Even reporting bugs helps.
ā
Cheers / Saludos,
Carlos E. R.
(from 13.1 x86_64 āBottleā at Telcontar)
Hi guys. okā¦
first of I didnt try it out on our ProxSys system itself.
We had 2 server running exactly the same stuff. But the one of them was just there for rendering preview files for the video files.
(Dont ask me why they used 2 server that support each 2 CPUs but they installed 2 servers with just one installed CPU on each of them, looks like someone just wanted to make some money, what a rip off)
It is absolutely the same copy of the other server. But this renderer didnt work for ages here( I just work 2 months here now )
Okā¦ I decided to do that kernel upgrade on the renderer as it was offline anyways and wasnt necessary to run the main aplication (that is web based)
For some freaking reason they didnt even install the wget packageā¦ grrā¦ luckily it has yast ā¦ usually it had to have it anyways.
But without a DVD of Linux 10.1 couldnt install any packages. So had to do it over a SD card and so on and onā¦
After all got to the point of updating the kernelā¦ before pressing enter just expected endless error messages.
But hey hoā¦ NOTHINGā¦ it did its job and I restarted the system andā¦ TADAAAAA GRUB already saw the whole 24GB and the kernel can see it too. free -m gives me the full 24GB
YES! I didnt see any malfunction so far.
So I am going to prepare that upgrade on the main application server now!
****ā¦ thanks a lot Miukuā¦
I will write back as soon I have updated the kernel.
In another Forum I got the hint, that I could even update the kernel to the latest build, as he said they dont really change the core so dramatically. So it should even work with that.
But of course that might be a bigger step and has without any doubts many more risks and side effects.
And he mentioned. Even with the bigsmp verison each CPU core can still only adress up to 4GB and not more, is that true?
Well as I have upgdared the server from one dual core xenon to now 2x quad core xenons that makes 8 cores that can adress up to 4GB each???.. hmmā¦ well I have to see how that works.
Today I am backing up the system first.
You shouldnāt believe everything you read on the net, Iāll just leave it at that.
Also updating to āthe latest kernelā just like that is not going to work - Iāll save you the trouble and tell you that if you try it, you will end up with a system that will not boot.
Well yeah, sounded kind of weird to me too. So thats why I had to ask here again. And as I can see the upgraded bigsmp version runs good so far. I wont play any stupid games.
Never change a running systemā¦ well I had to change it a little at least. Was a pain in the ass working with 2 useres on that browser based application.
if you do, try to stick to the versions āsupportedā - even though it isnāt really supported. :3
In theory it might be possible to compile a new kernel on the system (although Iād wager the old gcc might choke with a new kernel - Iāll have to try that someday ) and make an rpm out of it but letās face the facts: itās old as sin and youāre better off upgrading the entire system if that is possible sometime in the future, however such a complex task depends heavily on how the application running on the server was designed and what magical dependencies it might have.
I would advice against any changes to the system if they arenāt placing the system in any danger (such as being outside a firewall).
I think I got it wrong, he obviously meant not each single core but each single process that is being opened supports only 4GBā¦ well no idea if thats true either. Have to see it first.
Well lets forget about that kernel thing no need for that. I dont know for how long we will use this system. Depends on when it crashes But even if it crashes I have enough material here to reinstall a new server with the same configuration. So we can survive a few more years anyways.
About the security topic. People here have no freaking idea what security even means. Just when I arrived here it was a whole mess, everything. From the network to anyhting else, you dont want to see the server room. Looks worse than the jungle of amazon hehehe
At least this system works inside our network and has no direct connection to the internet. Even tho I dont want to talk about all the possibilities you can enter it anyways.
But I have to set priorities now, cannot do everything. And learning by doing myself as well. Each company especially this one has very unique and custom setting from hardware to software, was new to me too.
On 2015-01-29 13:26, VS D Media wrote:
>
> I think I got it wrong, he obviously meant not each single core but each
> single process that is being opened supports only 4GBā¦ well no idea if
> thats true either. Have to see it first.
Ā«With PAE, IA-32 architecture is augmented with additional address lines
used to select the additional memory, so physical address size increases
from 32 bits to 36 bits. This increases the physical memory addressable
by the system from 4 GB to 64 GB. The 32-bit size of the virtual address
is not changed, so regular application software continues to use
instructions with 32-bit addresses and (in a flat memory model) is
limited to 4 gigabytes of virtual address space. Operating systems
supporting this mode use page tables to map the regular 4 GB address
space into the physical memory, which, depending on the operating
system, may be as big as 64 GB. The mapping is typically applied
separately for each process, so that the extra memory is useful even
though no single regular application can access it all simultaneously.Ā»
ā
Cheers / Saludos,
Carlos E. R.
(from 13.1 x86_64 āBottleā at Telcontar)
On 01/29/2015 04:16 AM, VS D Media wrote:
>
> In another Forum I got the hint, that I could even update the kernel to
> the latest build, as he said they dont really change the core so
> dramatically. So it should even work with that.
> But of course that might be a bigger step and has without any doubts
> many more risks and side effects.
>
> And he mentioned. Even with the bigsmp verison each CPU core can still
> only adress up to 4GB and not more, is that true?
> Well as I have upgdared the server from one dual core xenon to now 2x
> quad core xenons that makes 8 cores that can adress up to 4GB each???..
> hmmā¦ well I have to see how that works.
> Today I am backing up the system first.
>
>
As it now seems you have a spare server perhaps you can upgrade that to
a newer OS version to test if the software will still work. Just an idea.
Miuku, just wanted to thank you over again!
I am not even sure if it really runs so much faster, but at least the rendering and the overall speed has increased. The software itself is still a piece of ****. But now the workflow is a bit better than before.
No worries, happy to help. Let us know if you need help with something else.
If you donāt mind me asking, where do you guys have your shop setup (NA, Europe, Asia?) and what kind of stuff do you do? It seems very interesting and specific work.
Wow. 16 GB in use already, when you had only 2 available. It appears
that you really needed that memory. āfree -hā may say what it is used
for. My guess is that a lot will be in disk cache.
ā
Cheers / Saludos,
Carlos E. R.
(from 13.1 x86_64 āBottleā at Telcontar)
and yes its mostly in cacheā¦ I start to think there is some kind of bug or what ever. It is faster. But I dont really know what all the ram is used for. The database is around 500MB compressed. Even uncompressed it shouldnt be that huge. Probably that Java (Jboss) version is too old and has many bugsā¦ no clue
total used free shared buffers cachedMem:
24367 22462 1904 0 0 21418
-/+ buffers/cache: 1044 23323
Swap: 1043 25 1017
btw
We are in Europe, and this is a small TV channel. The ProxSys system itself is used for various things. Like a normal video archive system. This version we have here has more modules included that offers a scheduling option for programing a normal TV channel. Including to that we have a special Playout system that is connected to the ProxSys too. The ProxSys software then generates a file that the playout can read (the dayily scheduling file) As well the Playout system is connected to another server that is a CG (character/overlay/lower third) generator.
We have more different hardware stuff (vor broadcasting applications)
Note the Linux hates unused memory so it will fill it up with cache. Remember that you are currently running 32 bit so though the system can now address a lot more each process is limited to under 4gig memory allocation.
makes sense yeah but I am watching the process via $ top and yes Java is the one who uses the biggest part of the ram but still too less. I think I saw some settings here on the system relating that JBoss Java server and it was limited to a specific amount of RAM. I think I will try to put a bigger RAM limit and see how it works. There are several processes that should be able to adress more ram.
does anyone know why the database can be that slow? is it only raw CPU power that is necessaryā¦ I cannot imagine that.
No 32 bit means 32 bit no 32 bit application can address more then 4gig no matter what you do. The PAE trick is a bank switch scheme whic allows the system to manage more then 4 gig.
Only way around it is to go full 64 bit.
Note you can still run 32 bit apps on a 64 bit but the application is still limited to 4 gig address space
I know that it is limited to 4gigs but what makes me a bit confused is. There is the main process JAVA and it has around 700MBā¦ far away from 4gigs.
And there are several other processes that are related to the database that are all around 500MB each. The database opens and closes processes all the time, none of them exceed more than ~500MB and I remember that I saw some setting that limited the Java or Database application to max 512MB. Just dont remember where I saw it
I am working and watching the processes all the time now and didnt see a single one goes over 1gig.
