High load average on fresh install - 9.3

[Carlos Capriotti]

Hello all.

I've posted something similar under "performance", but now I am more convinced this might be a real bug.

Here is the scenario: Two Dell PE T610s working side by side.

"Problem-server" is a freshly installed FreeNAS 9.3 with the latest update applied, but nothing else configured, except for networking, in order to gain access to the internet (for updates).

"Control-server"is its brother with a very similar but not identical configuration, behaving well, as expected.

Problem-server shows an average load on "top", "system processes" and on the reporting graphs, which is consistent and always between 0.6 and 0.7. Top and "processes" show nothing using the CPU.

The only difference that could matter between problem and control servers is that "problem" server is generation one, and control is generation 2. This may not be meaningful, since problem server was tested with CentOS 6.5 and the symptom was not present.

Now, before getting too serious about this as an official bug report, I thought of discussing it first.

Let me know what specifics you would like to know about this system.

Thank you.

 

[DifferentStrokes]

Let me know what specifics you would like to know about this system.

How about a little more information on:
Processor clock speed
Number of processor cores
Hyperthreading?
RAM
Jails?
VMs?

 

[Carlos Capriotti]

By all means, let's get the ball rolling.

2 x quad core (I thiiiiiiinnnk they are 2.4 GHz. Sorry, cannot remember form the top of my head but sounds pretty accurate)
Hyperthread YES
24 GB RAM

Absolutely no other configuration of any type, except for Intel NIC (off board) for Internet access, thus, no jails or VMs.

Both servers with same BIOS settings.

Problem-system has already been fully configured before, displaying the same symptom.

 

[DifferentStrokes]

Does

Code:

ps -A

at the command line reveal anything out of the ordinary?

 

[Carlos Capriotti]

Sorry for the delay. Lost contact with the server yesterday.

Comparing the output of both servers, no, I can tell you that except for the nfs related stuff of the production server, and one or two other bits also linked to production, both ps -A seem to have the same content.

 

[enemy85]

just a consideration: the value of 1.0 should be the 100% use of a single core, so in your case the 100% of CPU use should be equal to a value of 8.0

 

[Carlos Capriotti]

You are right, but, this system is using an entire processor (ok, a bit of a stretch, but bear with me) doing nothing... ?
Because of hyperthread, this system reports 16 cores, so, in fact it is using a bit over 1/4 of one core, fair enough, but, doing nothing ?

It is a fresh installed system, with virtually nothing running on it; and other servers will report the expected 0.0x instead of a constant 0.7. Other operating systems on this server also correctly report 0.0x.

People use this expression quite frequently, so I am going to paraphrase: this system is using an extra order of magnitude of resources to do... noting ?

Feels like a bug, but where ? Well, that's why i am here.

Let's add evidence, shall we ? This is the problem system. Constant resource utilization.

Next is the control system, which pretty much shows the expected (and average) use on a normal, fairly idle system.

 

[Carlos Capriotti]

@FreeNAS team, @iX, @Moderators, I'd need a hand here.

I've tested FreeBSD 9.4 AMD64 LiveCD (well, mem stick) on this system and the problem is present there.

How would you like me to proceed ?