[Linux-PowerEdge] 15 or 20 package_power_limit events per second on an R320

Todd Lyons tlyons at ivenue.com
Tue Jul 2 07:51:48 CDT 2013


On Tue, Mar 19, 2013 at 11:27 PM, Paul Warner <paul.warner at cyaninc.com> wrote:
> Based on this input as well as some advice from our Dell rep we disabled
> "cstates" and "c1e".  This seemed to clear things up.  I guess from what I
> read about what "c1e" does and what the power_limit evens from the kernel
> suggest, it probably means we were stuck with too low a voltage to the CPU I
> guess?  So not enough power to run at the proper rate?

We have some R320's that are doing this.  It's occurred once on both
CentOS 6.x (kernel 2.6.32-279.el6.x86_64) and Debian 6.0 (kernel
2.6.32-16-pve, part of Proxmox virtualization platform).  We can
always tell that it's happening because a simple dd command results in
extremely low I/O throughput:

# dd if=/dev/zero of=/tmp/dummy.txt bs=1M count=200
200+0 records in
200+0 records out
209715200 bytes (210 MB) copied, 7.62787 s, 27.5 MB/s

A software reboot does not resolve the error; it requires cycling the
power.  After the hard reboot, things are back to normal:

# dd if=/dev/zero of=/tmp/dummy.txt bs=1M count=200
200+0 records in
200+0 records out
209715200 bytes (210 MB) copied, 0.384195 s, 546 MB/s

> Any ways, it is better.  Thanks.  We plan to update the BIOS and re-enable
> and see if that still leaves us in a working state.  It sounds like "c1e" at
> least used to reduce idle power consumption significantly at zero
> performance impact, so it would be nice to have it enabled.

Paul, have you experienced any further issues?  What ended up being
your final resolution?  Did you change both C1E and C-States to
disabled?  Or did one or the other get rid of the issue?  Or did a
BIOS update fix everything?  What did disabling these power saving
modes do to your power consumption?

> On Wed, Mar 13, 2013 at 11:27 PM, <Srinivas_G_Gowda at dell.com> wrote:
>>
>> You are probably seeing this because you might have set "Active power
>> cap policy" in iDrac.

We didn't actually configure the iDrac at all.  I'll double check to
see what's there, but based on the above comments, it's the C1E and
C-States being on that are the root cause for us.

...Todd

--
The total budget at all receivers for solving senders' problems is $0.
 If you want them to accept your mail and manage it the way you want,
send it the way the spec says to. --John Levine



More information about the Linux-PowerEdge mailing list