OVH Community, your new community space.

disable speedstep?


raxxeh
25-08-2013, 04:12
Quote Originally Posted by rizuk
ok cheers had no nic problems with 3.10?
I'm not using 3.10, I'm using 3.9.4 on one server, and 2.6 on my proxmox host.

rizuk
25-08-2013, 03:46
ok cheers had no nic problems with 3.10?

raxxeh
25-08-2013, 03:44
Quote Originally Posted by rizuk
how did you remove speedstep on 3.10
you don't use OVH's kernels, that's how...

after 30min of pissing around and having nothing work i just swapped directly over to another kernel i use on a different server and it worked straight away

rizuk
25-08-2013, 03:41
Quote Originally Posted by raxxeh
I have processes that pop up occasionally that need max cpu for about 1.5 seconds at a time... it's faster if i run it at full rate at all times.

the extra power usage is insignificant if its sitting at full mhz/ghz, but using 0%
how did you remove speedstep on 3.10

raxxeh
25-08-2013, 03:25
Quote Originally Posted by Myatu
I presume that's because it will affect the datacenter's PUE if everyone started to disabled it. But does disabling speedstep make any difference in performance though? I have my doubts.
I have processes that pop up occasionally that need max cpu for about 1.5 seconds at a time... it's faster if i run it at full rate at all times.

the extra power usage is insignificant if its sitting at full mhz/ghz, but using 0%

JakeMS
24-08-2013, 16:09
Quote Originally Posted by Myatu
I presume that's because it will affect the datacenter's PUE if everyone started to disabled it. But does disabling speedstep make any difference in performance though? I have my doubts.
This entirely dependant on what the server is doing.

Some tasks will benefit from quicker response by the cpu (quicker frequency)

For example, a task which only pops up when it needs to for a few seconds will get a greater benefit than a task which runs full CPU all the time.

Where as a task which can use a lot of CPU power, with multiple cores (such as compiling applications, video editing etc) will be much better off with a CPU which has more cores with possibly lower frequencies more so than one that uses a single core and bursts from time to time as a CPU may not "switch" speeds quick enough, for the bursting task, compared to a task which runs for a longer duration where it has plenty of time to switch at will.

For example, I've seen some Apache/Memcached/PHP applications get a great benefit from not switching speeds (down from ~1.2s to ~300-~500ms).

Where as tasks which run longer, tend not to notice.

Myatu
24-08-2013, 16:04
I presume that's because it will affect the datacenter's PUE if everyone started to disabled it. But does disabling speedstep make any difference in performance though? I have my doubts.

raxxeh
24-08-2013, 01:21
Just a heads up; OVH's new 3.10 kernel disables almost all methods to turn off speed-step.

Ended up reverting a server to get around it.

rizuk
09-08-2013, 15:01
things i did

apt-get install linux-image-server
mkdir ~/ovh.d
mv /etc/grub.d/06_OVHkernel ~/ovh.d
update-grub

rebooted sorted was on 3.2.0.51
then tried to update it to 3.5
sudo apt-get install linux-image-generic-lts-quant
update-grub

didnt boot

so i after that i did try 3.8 no that didnt boot


menuentry 'Ubuntu, with Linux 3.8.0-27-generic' --class ubuntu --class gnu-linux --class gnu --class os {
menuentry 'Ubuntu, with Linux 3.8.0-27-generic (recovery mode)' --class ubuntu --class gnu-linux --class gnu --class os {
menuentry 'Ubuntu, with Linux 3.5.0-37-generic' --class ubuntu --class gnu-linux --class gnu --class os {
menuentry 'Ubuntu, with Linux 3.5.0-37-generic (recovery mode)' --class ubuntu --class gnu-linux --class gnu --class os {
menuentry 'Ubuntu, with Linux 3.2.0-51-generic' --class ubuntu --class gnu-linux --class gnu --class os {
menuentry 'Ubuntu, with Linux 3.2.0-51-generic (recovery mode)' --class ubuntu --class gnu-linux --class gnu --class os {
menuentry "Ubuntu 12.04, OVH kernel 3.8.13-xxxx-grs-ipv6-64" {


was trying to go back to Linux 3.2.0-51-generic as that was working set it to load that but no luck

JakeMS
09-08-2013, 11:13
Quote Originally Posted by rizuk
oh crap updated kernel to 3.8 now its not booting lolz
went into ovh loaded a kernel rescue that booted fine
went into grub changed it to the correct kernel updated grub

Kernel didnt load
What do your logs say?

is it a kernel panic? unable to find grub? can't start network? etc?

rizuk
09-08-2013, 02:37
oh crap updated kernel to 3.8 now its not booting lolz
went into ovh loaded a kernel rescue that booted fine
went into grub changed it to the correct kernel updated grub

Kernel didnt load

Myatu
08-08-2013, 23:29
Quote Originally Posted by rizuk
hey mate yeah thing is i did try that and it just booted me out of the ssh lol
There was a warning about that

JakeMS
08-08-2013, 23:24
Quote Originally Posted by rizuk
hey mate yeah thing is i did try that and it just booted me out of the ssh lol
which kernel is best do you think i have 3.2 atm

thanks to your other post i managed to change the ovh kernel
As long as you stay below 3.9 you can install the driver, so you should be okay :-)

Though if you fancy running CentOS or RHEL with stock kernels you could try and install this:

http://elrepo.org/tiki/kmod-r8168

I am unable to install or the above on my CentOS desktop as the stock kernel craps it's self so I run mainline which works fine. However, in the past I have used other packages from that repository on servers, such as kmod-tpe :-).

rizuk
08-08-2013, 23:06
hey mate yeah thing is i did try that and it just booted me out of the ssh lol
which kernel is best do you think i have 3.2 atm

thanks to your other post i managed to change the ovh kernel

Myatu
08-08-2013, 23:04
Quote Originally Posted by rizuk
[ 1.816807] r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
You didn't update the driver, as outlined in one of the other posts: http://forum.ovh.co.uk/showpost.php?p=47569&postcount=9

rizuk
08-08-2013, 22:58
[ 11.218188] r8169 0000:03:00.0: eth0: link down
[ 11.218193] r8169 0000:03:00.0: eth0: link down
[ 11.218869] ADDRCONF(NETDEV_UP): eth0: link is not ready

maybe the reason humm

rizuk
08-08-2013, 22:57
ok its not the cpu it has to do with the nic it only seems to crash when a lot of connections or over 100mbps is used

can you see anything in these logs


[ 1.732188] ata2.00: ATA-8: TOSHIBA DT01ACA200, MX4OABB0, max UDMA/133
[ 1.732196] ata2.00: 3907029168 sectors, multi 16: LBA48 NCQ (depth 31/32), AA
[ 1.732946] ACPI Error: [DSSP] Namespace lookup failure, AE_NOT_FOUND (20110623/psargs-359)
[ 1.732960] ACPI Error: Method parse/execution failed [\_SB_.PCI0.SAT0.SPT0._GTF] (Node ffff8807ebc87730), AE_NOT_FOUND (20110623/psparse-536)
[ 1.733272] ata1.00: configured for UDMA/133
[ 1.733376] ACPI Error: [DSSP] Namespace lookup failure, AE_NOT_FOUND (20110623/psargs-359)
[ 1.733382] ACPI Error: Method parse/execution failed [\_SB_.PCI0.SAT0.SPT1._GTF] (Node ffff8807ebc877a8), AE_NOT_FOUND (20110623/psparse-536)
[ 1.733412] scsi 0:0:0:0: Direct-Access ATA TOSHIBA DT01ACA2 MX4O PQ: 0 ANSI: 5
[ 1.733491] sd 0:0:0:0: Attached scsi generic sg0 type 0
[ 1.733520] sd 0:0:0:0: [sda] 3907029168 512-byte logical blocks: (2.00 TB/1.81 TiB)
[ 1.733524] sd 0:0:0:0: [sda] 4096-byte physical blocks
[ 1.733562] sd 0:0:0:0: [sda] Write Protect is off
[ 1.733565] sd 0:0:0:0: [sda] Mode Sense: 00 3a 00 00
[ 1.733592] sd 0:0:0:0: [sda] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[ 1.733687] ata2.00: configured for UDMA/133
[ 1.733836] scsi 1:0:0:0: Direct-Access ATA TOSHIBA DT01ACA2 MX4O PQ: 0 ANSI: 5
[ 1.733915] sd 1:0:0:0: Attached scsi generic sg1 type 0
[ 1.733930] sd 1:0:0:0: [sdb] 3907029168 512-byte logical blocks: (2.00 TB/1.81 TiB)
[ 1.733934] sd 1:0:0:0: [sdb] 4096-byte physical blocks
[ 1.733983] sd 1:0:0:0: [sdb] Write Protect is off
[ 1.733995] sd 1:0:0:0: [sdb] Mode Sense: 00 3a 00 00
[ 1.734022] sd 1:0:0:0: [sdb] Write cache: enabled, read cache: enabled, doesn't support DPO or FUA
[ 1.742727] usb 1-1: new high-speed USB device number 2 using ehci_hcd
[ 1.767984] sda: sda1 sda2 sda3 sda4
[ 1.768320] sd 0:0:0:0: [sda] Attached SCSI disk
[ 1.782652] sdb: sdb1 sdb2 sdb3 sdb4
[ 1.783220] sd 1:0:0:0: [sdb] Attached SCSI disk
[ 1.784076] Freeing unused kernel memory: 924k freed
[ 1.784143] Write protecting the kernel read-only data: 12288k
[ 1.786803] Freeing unused kernel memory: 1592k freed
[ 1.788835] Freeing unused kernel memory: 1188k freed
[ 1.798766] udevd[121]: starting version 175
[ 1.807795] Btrfs loaded
[ 1.808781] md: linear personality registered for level -1
[ 1.810095] md: multipath personality registered for level -4
[ 1.811287] md: raid0 personality registered for level 0
[ 1.812483] md: raid1 personality registered for level 1
[ 1.813853] async_tx: api initialized (async)
[ 1.816807] r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
[ 1.816827] r8169 0000:03:00.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[ 1.816867] r8169 0000:03:00.0: setting latency timer to 64
[ 1.816950] r8169 0000:03:00.0: irq 45 for MSI/MSI-X
[ 1.817759] r8169 0000:03:00.0: eth0: RTL8168f/8111f at 0xffffc9000303a000, 74:d0:2b:27:d6:0e, XID 08000880 IRQ 45
[ 1.817765] r8169 0000:03:00.0: eth0: jumbo features [frames: 9200 bytes, tx checksumming: ko]
[ 1.875333] hub 1-1:1.0: USB hub found
[ 1.875440] hub 1-1:1.0: 6 ports detected
[ 1.878648] raid6: int64x1 4280 MB/s
[ 1.946604] raid6: int64x2 4756 MB/s
[ 1.960650] md: bind
[ 1.986588] usb 2-1: new high-speed USB device number 2 using ehci_hcd
[ 2.014548] raid6: int64x4 3960 MB/s
[ 2.082504] raid6: int64x8 3217 MB/s
[ 2.119215] hub 2-1:1.0: USB hub found
[ 2.119308] hub 2-1:1.0: 8 ports detected
[ 2.150465] raid6: sse2x1 10195 MB/s
[ 2.218423] raid6: sse2x2 12670 MB/s
[ 2.286382] raid6: sse2x4 14700 MB/s
[ 2.286384] raid6: using algorithm sse2x4 (14700 MB/s)
[ 2.286408] Refined TSC clocksource calibration: 3399.938 MHz.
[ 2.286420] Switching to clocksource tsc
[ 2.286458] md: bind
[ 2.287012] xor: automatically using best checksumming function: generic_sse
[ 2.287751] md: bind
[ 2.288618] bio: create slab at 1
[ 2.288675] md/raid1:md2: active with 2 out of 2 mirrors
[ 2.288690] md2: detected capacity change from 0 to 1978380779520
[ 2.289359] md: bind
[ 2.289504] md2: unknown partition table
[ 2.290153] md/raid1:md1: not clean -- starting background reconstruction
[ 2.290157] md/raid1:md1: active with 2 out of 2 mirrors
[ 2.290168] md1: detected capacity change from 0 to 21474770944
[ 2.290193] md: resync of RAID array md1
[ 2.290197] md: minimum _guaranteed_ speed: 1000 KB/sec/disk.
[ 2.290201] md: using maximum available idle IO bandwidth (but not more than 200000 KB/sec) for resync.
[ 2.290206] md: using 128k window, over a total of 20971456k.
[ 2.306353] generic_sse: 16691.000 MB/sec
[ 2.306356] xor: using function: generic_sse (16691.000 MB/sec)
[ 2.306996] md: raid6 personality registered for level 6
[ 2.306999] md: raid5 personality registered for level 5
[ 2.307000] md: raid4 personality registered for level 4
[ 2.309221] md: raid10 personality registered for level 10
[ 2.328202] md1: unknown partition table
[ 7.879684] EXT4-fs (md1): INFO: recovery required on readonly filesystem
[ 7.879688] EXT4-fs (md1): write access will be enabled during recovery
[ 8.696831] EXT4-fs (md1): orphan cleanup on readonly fs
[ 8.696838] EXT4-fs (md1): ext4_orphan_cleanup: deleting unreferenced inode 655371
[ 8.696853] EXT4-fs (md1): ext4_orphan_cleanup: deleting unreferenced inode 655370
[ 8.696857] EXT4-fs (md1): ext4_orphan_cleanup: deleting unreferenced inode 655369
[ 8.696860] EXT4-fs (md1): ext4_orphan_cleanup: deleting unreferenced inode 655368
[ 8.696863] EXT4-fs (md1): ext4_orphan_cleanup: deleting unreferenced inode 655367
[ 8.696867] EXT4-fs (md1): 5 orphan inodes deleted
[ 8.696868] EXT4-fs (md1): recovery complete
[ 8.757203] EXT4-fs (md1): mounted filesystem with ordered data mode. Opts: (null)
[ 9.720934] Adding 525308k swap on /dev/sda3. Priority:-1 extents:1 across:525308k
[ 9.721956] Adding 525308k swap on /dev/sdb3. Priority:-2 extents:1 across:525308k
[ 9.790349] ADDRCONF(NETDEV_UP): eth0: link is not ready
[ 9.826401] udevd[405]: starting version 175
[ 9.998909] EXT4-fs (md1): re-mounted. Opts: errors=remount-ro
[ 10.700823] mei: module is from the staging directory, the quality is unknown, you have been warned.
[ 10.701013] mei 0000:00:16.0: PCI INT A -> GSI 21 (level, low) -> IRQ 21
[ 10.701018] mei 0000:00:16.0: setting latency timer to 64
[ 10.701061] mei 0000:00:16.0: irq 46 for MSI/MSI-X
[ 10.710521] parport_pc 00:09: reported by Plug and Play ACPI
[ 10.710585] parport0: PC-style at 0x378, irq 5 [PCSPP]
[ 10.714433] wmi: Mapper loaded
[ 10.775251] asus_wmi: ASUS WMI generic driver loaded
[ 10.776168] asus_wmi: Initialization: 0x0
[ 10.776181] asus_wmi: BIOS WMI version: 0.9
[ 10.776202] asus_wmi: SFUN value: 0x0
[ 10.776345] input: Eee PC WMI hotkeys as /devices/platform/eeepc-wmi/input/input3
[ 10.776754] asus_wmi: Backlight controlled by ACPI video driver
[ 10.807930] ppdev: user-space parallel port driver
[ 10.870572] lp0: using parport0 (interrupt-driven).
[ 10.889947] [drm] Initialized drm 1.1.0 20060810
[ 10.906404] type=1400 audit(1375824457.902:2): apparmor="STATUS" operation="profile_load" name="/sbin/dhclient" pid=640 comm="apparmor_parser"
[ 10.906410] type=1400 audit(1375824457.902:3): apparmor="STATUS" operation="profile_replace" name="/sbin/dhclient" pid=639 comm="apparmor_parser"
[ 10.906678] type=1400 audit(1375824457.902:4): apparmor="STATUS" operation="profile_load" name="/usr/lib/NetworkManager/nm-dhcp-client.action" pid=640 comm="apparmor_par ser"
[ 10.906683] type=1400 audit(1375824457.902:5): apparmor="STATUS" operation="profile_replace" name="/usr/lib/NetworkManager/nm-dhcp-client.action" pid=639 comm="apparmor_ parser"
[ 10.906836] type=1400 audit(1375824457.902:6): apparmor="STATUS" operation="profile_load" name="/usr/lib/connman/scripts/dhclient-script" pid=640 comm="apparmor_parser"
[ 10.906842] type=1400 audit(1375824457.902:7): apparmor="STATUS" operation="profile_replace" name="/usr/lib/connman/scripts/dhclient-script" pid=639 comm="apparmor_parse r"
[ 10.909349] EXT4-fs (md2): mounted filesystem with ordered data mode. Opts: (null)
[ 11.095522] i915 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
[ 11.095526] i915 0000:00:02.0: setting latency timer to 64
[ 11.141331] mtrr: type mismatch for e0000000,10000000 old: write-back new: write-combining
[ 11.141332] [drm] MTRR allocation failed. Graphics performance may suffer.
[ 11.141534] i915 0000:00:02.0: irq 47 for MSI/MSI-X
[ 11.141536] [drm] Supports vblank timestamp caching Rev 1 (10.10.2010).
[ 11.141537] [drm] Driver supports precise vblank timestamp query.
[ 11.141578] [drm:intel_dsm_platform_mux_info] *ERROR* MUX INFO call failed
[ 11.141867] vgaarb: device changed decodes: PCI:0000:00:02.0,olddecodes=io+mem,decodes=io+memw ns=io+mem
[ 11.218188] r8169 0000:03:00.0: eth0: link down
[ 11.218193] r8169 0000:03:00.0: eth0: link down
[ 11.218869] ADDRCONF(NETDEV_UP): eth0: link is not ready
[ 11.479405] fbcon: inteldrmfb (fb0) is primary device
[ 11.647934] Console: switching to colour frame buffer device 128x48
[ 11.649067] fb0: inteldrmfb frame buffer device
[ 11.649068] drm: registered panic notifier
[ 11.724977] acpi device:42: registered as cooling_device13
[ 11.725100] input: Video Bus as /devices/LNXSYSTM:00/LNXSYBUS:00/PNP0A08:00/LNXVIDEO:00/input/input4
[ 11.725190] ACPI: Video Device [GFX0] (multi-head: yes rom: no post: no)
[ 11.725224] [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0
[ 12.260787] init: failsafe main process (693) killed by TERM signal
[ 12.521510] type=1400 audit(1375824459.518:8): apparmor="STATUS" operation="profile_replace" name="/sbin/dhclient" pid=908 comm="apparmor_parser"
[ 12.521751] type=1400 audit(1375824459.518:9): apparmor="STATUS" operation="profile_replace" name="/usr/lib/NetworkManager/nm-dhcp-client.action" pid=908 comm="apparmor_ parser"
[ 12.521882] type=1400 audit(1375824459.518:10): apparmor="STATUS" operation="profile_replace" name="/usr/lib/connman/scripts/dhclient-script" pid=908 comm="apparmor_pars er"
[ 12.546526] type=1400 audit(1375824459.542:11): apparmor="STATUS" operation="profile_load" name="/usr/sbin/named" pid=910 comm="apparmor_parser"

Andy
07-08-2013, 00:06
Run the server in netboot rescue and run all of the tests. If it fails those then you know it's a H/W issue. If not, then it's more likely something you've done software wise causing it. By the way, OVH will only ask you to do the same thing so beat them to it and do it first.

JakeMS
06-08-2013, 23:06
If you're referring to the Realtek NIC, then the way is simple:

When the realtek driver issue occurs it's network only. Everything else works normally. (At least, that's how it always is for me).

Otherwise it's something else.

rizuk
06-08-2013, 22:52
crash crash crash ovh sort it out something is wrong ;/

how would i know if its the nic or cpu that issue

Myatu
06-08-2013, 21:29
Install lm-sensors ("apt-get install lm-sensors" on Debian/Ubuntu) and run "sensors-detect", following its instructions (mostly consisting of pressing ENTER except at the very last step). You can then check the temperatures on the server using "sensors".

JakeMS
06-08-2013, 21:17
Quote Originally Posted by Andy
If set incorrectly then the CPU can cause crashing. When I messed around with my own settings on my laptop I had random crash issues. Unless it's causing you an actual problem I'd leave it alone. Intel Speedstep is good at what it does on it's own without intervention.

Would you rather have reliability or a script that runs a little bit quicker? Stability all the way for me.
Unless you're trying to tell the cpu to step up to something it is not designed to do, it should never crash unless it's got a cooling problem (most laptops do)

If you disable cpuspeed entirely the cpu should run at stock speed without any problems what so ever, and if indeed it does have problems that indicates a hardware, or cooling problem as the cpu should by design be able to run at full speed without issues.

rizuk
06-08-2013, 21:16
the script does not effect the overall cpu usage so the idel speeds do not change= High script usage

Andy
06-08-2013, 21:07
If set incorrectly then the CPU can cause crashing. When I messed around with my own settings on my laptop I had random crash issues. Unless it's causing you an actual problem I'd leave it alone. Intel Speedstep is good at what it does on it's own without intervention.

Would you rather have reliability or a script that runs a little bit quicker? Stability all the way for me.

rizuk
06-08-2013, 21:05
my server is now constantly crashing

The incident detected on 2013-08-06 20:48:04 is now
resolved. We have closed the thread. However, if you want
to re-open it, please use the web interface.
From : Monitoring OVH For: l6-ovh Hide the message
Date: 2013-08-06 20:52:27


The intervention on ns2.ovh.net has been completed.

This operation was closed at 2013-08-06 20:52:27

Here are the details of this operation:
HARD Reboot
Date 2013-08-06 20:49:45, gael.van-respaille made HARD
Reboot:
Here are the details of the operation performed:
No information on the screen ("black screen"). No response
to keyboard.

Actions:
Restart the server hardware.

result:
Boot OK. Server can 'login'. Ping OK, services started.

first i thought it was the nic driver now I'm not so sure

I'm sure this must be to do when i change cpu speeds any fix for this?

JakeMS
06-08-2013, 13:41
For CentOS / Redhat users who wish to disable speedstepping:

service cpuspeed stop
chkconfig cpuspeed off

This will stop the stepping being enabled at all, and thus cpu will run at whatever the BIOS tells it to (typically full speed).

elcct
06-08-2013, 13:37
I had the same problem with my MG server.

Command that solved it:

for CPUFREQ in /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor; do [ -f $CPUFREQ ] || continue; echo -n performance > $CPUFREQ; done

Reboot was not necessary

raxxeh
05-08-2013, 23:44
Quote Originally Posted by rizuk
cheers

yeah its working wished you could set it to 3.8 fully

how does that work out you stress all cores but you're only getting 3.5ghz
you stress one thread all cores go to 3.8ghz
Read what the program tells you. it is explained in the screenshot, turbo isn't 3.8 on all cores.

rizuk
05-08-2013, 23:39
cheers

yeah its working wished you could set it to 3.8 fully

how does that work out you stress all cores but you're only getting 3.5ghz
you stress one thread all cores go to 3.8ghz

raxxeh
05-08-2013, 23:34
Quote Originally Posted by rizuk
what program is that

this what the problem is program i use does not stress it to hit 3.8 i would need to stress one thread for it to hit 3.8
Quote Originally Posted by raxxeh
https://code.google.com/p/i7z/

Try using this. I am experimenting with my mSP now, will let you know if i find a way to get it over 3400.

first image: 8 thread sysbench

second image: 1 thread sysbench.

turbo is working.

rizuk
05-08-2013, 23:22
what program is that

this what the problem is program i use does not stress it to hit 3.8 i would need to stress one thread for it to hit 3.8

raxxeh
05-08-2013, 23:18
.....

rizuk
05-08-2013, 23:15
let me know im looking into it the cpu should boost to 3.8ghz

rizuk
05-08-2013, 23:13
cheers so i can set it to 3.4ghz but the program does not stress the cpu
so when i run a benchmark on one thread it will hit 3.8ghz boost mode

allowing the program to use less cpu

raxxeh
05-08-2013, 23:13
https://code.google.com/p/i7z/

Try using this. I am experimenting with my mSP now, will let you know if i find a way to get it over 3400.

raxxeh
05-08-2013, 23:08
Quote Originally Posted by rizuk
what about when the server restarts mate would it go back to idel speeds
not that much of a linux geek wished i could disable all power states ;/

I would need to find out about this kernel

raxx do you know if you can run the cpu in full boost mode all the time? 3.8ghz
rc.local is executed on every single reboot, what I posted is permanant.

I don't think you can run 3.8ghz on these chips, have a feeling OVH disabled it.

none the less; you could set it statically with:

cpufreq-set --freq 3.8GHz -c 0

rizuk
05-08-2013, 23:02
what about when the server restarts mate would it go back to idel speeds
not that much of a linux geek wished i could disable all power states ;/

I would need to find out about this kernel

raxx do you know if you can run the cpu in full boost mode all the time? 3.8ghz

raxxeh
05-08-2013, 23:00
Quote Originally Posted by Thelen
need custom kernel to fix this, or to run a command line script which i'll find out from a friend shortly.
install 'cpufrequtils'

run cpufreq-info

you'll see most of them will be idle, to set them to lock at max run

cpufreq-set -g performance -c #

for an 8 core machine, to set them all to 3.4 on the msp it's:

cpufreq-set -g performance -c 0
cpufreq-set -g performance -c 1
cpufreq-set -g performance -c 2
cpufreq-set -g performance -c 3
cpufreq-set -g performance -c 4
cpufreq-set -g performance -c 5
cpufreq-set -g performance -c 6
cpufreq-set -g performance -c 7

you can recompile kernel and disable the other power states, but i find its easier to just dump that in rc.local.

Myatu
05-08-2013, 22:58
Setting "/sys/devices/system/cpu/cpu*/cpufreq/scaling_governor" is not permanent, and will be reset upon reboot.

rizuk
05-08-2013, 22:41
humm after the reboot Speed: 1600.000 MHz again
wonder if ovh can disable it

rizuk
05-08-2013, 22:06
thanks it was just down for about 5mins so i rebooted and it was still down for about 5mins and it came back on will check logs

Myatu
05-08-2013, 21:59
Has it crashed, but came back online, or is it still down?

If it is still down, a technician will notice this and perform an check on the system to see what's going on (provided you have not disabled OVH's active check). Just let them do their bit if this is the case.

If it has come back up, check if the uptime has changed (to see if it rebooted, or if it was just a network fault like the issue with the switch). Then check your /var/log/syslog, which should indicate when the server was rebooted and what happened before that (which would hopefully give an indication as to the cause).

rizuk
05-08-2013, 21:54
hum my server just crashed not sure if it was to do with http://status.ovh.co.uk/?do=details&id=5157

beucase that was at 5pm

might be where i changed cpu speeds?

Myatu
05-08-2013, 16:48
Looks like it may be a small bug with rounding numbers. But yeah, you've got it set to "performance" I see.

rizuk
05-08-2013, 16:13
thank you looks to be forced to 3.4ghz now


not sure about turbo boost


do you know why it displays the max frequency twice

Myatu
05-08-2013, 16:06
Check in your "/sys/devices/system/cpu".

For example, "/sys/devices/system/cpu/cpu*/cpufreq/scaling_governor" is often set by an "ondemand" init script (see if its in your /etc/init.d/ or /etc/init). This you can force to "performance" at all times.

You can also tweak "/sys/devices/system/cpu/cpufreq/ondemand/up_threshold" and other bits and bobs there. If you have "cpufreqd" or "cpuspeed" running, then you'll have to kill those first.

Have fun!

rizuk
05-08-2013, 15:55
ok cheers thelen my mate set the cpu to 3.4ghz no more 1.6ghz now
seems to lower the usage but still benching takes it down a few more percent

i think its because when i do a bench turbo mode will kick it to 3.8ghz on all cores?

Thelen
05-08-2013, 14:46
need custom kernel to fix this, or to run a command line script which i'll find out from a friend shortly.

rizuk
05-08-2013, 13:19
yeah well this program i use only uses one core talking of program cpu usage
not the full system

4/8threads

the program

as you can see

when i bench one thread to boost all cores to 3.4ghz this is the program cpu usage


i stop the bench and the core goes back to 1.6ghz as you can see the rise in usage



more connections more usage


its also should boost to 3.8ghz but under linux i can only see it hitting max 3.4

Andy
05-08-2013, 13:16
How many cores has it got? It won't move from 1.6GHz if the CPU isn't under enough load. So in actual fact it probably isn't causing any issues at all. If your CPU was hitting 100% then I would see it as an issue.

25% sounds like 1 core out of 4 being maxed out but this should still make it step up the frequency.

rizuk
05-08-2013, 13:08
Would ovh disable speedstep for me ? i use a few programs and it effects it
the cpu never moves from 1.6ghz when i run python app which is using 25 percent of its program cpu usage

I ran a bench on one thread to boost all cores to 3.4ghz the python program i was running cpu usage went down to 3 percent