After update of Fog Trunk - PHP Errors

[Tom Elliott]

@RipAU errors should be no more. I can only hope the cpu will settle down as it doesn’t have to try writing the error to the log as well.

[RipAU]

Ok, Just did an update to the latest trunk (in fog version its listed as 7306)
I have lost the graph that lists the amount of storage left and when clicking on the name it has access denied and wrong user name and password listed. I’m also receiving the following errors in HTTPD (different ones this time )

[Tue Apr 26 11:13:35.131109 2016] [:error] [pid 1499] [client 10.254.1.59:55287] PHP Fatal error:  Uncaught Error: Using $this when not in object context in /var/www/html/fog/lib/fog/fogpage.class.php:743\nStack trace:\n#0 /var/www/html/fog/commons/init.php(141): FOGPage::authorize()\n#1 /var/www/html/fog/commons/base.inc.php(10): require_once('/var/www/html/f...')\n#2 /var/www/html/fog/management/index.php(2): require('/var/www/html/f...')\n#3 {main}\n  thrown in /var/www/html/fog/lib/fog/fogpage.class.php on line 743

CPU usage is still hovering very high.

Thanks again for the help

[RipAU]

I decided to check some fog.log files on one of my hosts and I am also seeing this.

26/04/2016 11:27 AM Client-Info Version: 0.9.12
 26/04/2016 11:27 AM Middleware::Communication URL: http://fog2/fog/management/other/ssl/srvpublic.crt
 26/04/2016 11:27 AM Data::RSA FOG Server CA cert found
 26/04/2016 11:27 AM Middleware::Authentication Cert OK
 26/04/2016 11:27 AM Middleware::Communication POST URL: http://fog2/fog/management/index.php?sub=authorize
 26/04/2016 11:27 AM Middleware::Communication ERROR: Failed to POST data
 26/04/2016 11:27 AM Middleware::Communication ERROR: The remote server returned an error: (500) Internal Server Error.
 26/04/2016 11:27 AM Middleware::Authentication ERROR: Could not authenticate
 26/04/2016 11:27 AM Middleware::Authentication ERROR: Object reference not set to an instance of an object.
 26/04/2016 11:27 AM Service Sleeping for 120 seconds```

[Wayne Workman]

install apachetop and then run it.
apachetop

It will give you the request count of every file served by Apache, and the amount of data transmitted. This will help us at least narrow down exactly what file is causing your problems.

sample output:

[RipAU]

Just checked then.

I’m not sure I can see an issue? with the following.

last hit: 03:35:45         atop runtime:  0 days, 00:02:05             03:35:46
All:         2043 reqs (  16.3/sec)        467.5K ( 3829.8B/sec)     234.3B/req
2xx:    1817 (88.9%) 3xx:      52 ( 2.5%) 4xx:     2 ( 0.1%) 5xx:   172 ( 8.4%)
R ( 30s):     476 reqs (  15.9/sec)        116.9K ( 3990.2B/sec)     251.5B/req
2xx:     424 (89.1%) 3xx:       6 ( 1.3%) 4xx:     0 ( 0.0%) 5xx:    46 ( 9.7%)

 REQS REQ/S    KB KB/S URL
  144  4.80   0.7  0.0*/fog/service/servicemodule-active.php
   94  3.13   4.1  0.1 /fog/management/index.php
   66  2.20   2.4  0.1 /fog/status/bandwidth.php
   49  1.63  80.3  2.7 /fog/management/other/ssl/srvpublic.crt
   43  1.43   1.5  0.0 /fog/service/Printers.php
   14  0.48   0.1  0.0 /fog/service/greenfog.php
   13  0.43  27.5  0.9 /fog/service/printerlisting.php
   13  0.43   0.1  0.0 /fog/service/snapins.checkin.php
   11  0.37   0.1  0.0 /fog/service/hostname.php
   10  0.36   0.0  0.0 /fog/service/jobs.php
   10  0.36   0.1  0.0 /fog/service/getversion.php
    6  0.21   0.0  0.0 /fog/index.php
    3  0.10   0.0  0.0 *

looks like most requests are being sent to servicemodule-active.php ?

[RipAU]

Hi, Guys.

As a bit of a hopeful move I updated to the latest SVN update (FOG Version: 7330)
I’ve found the web interface near unusable because of the load that is now being put on the server(Huge number of httpd processes are now being spawned taking up most of the CPU and quite allot of mysqld processes as well)

I’ll roll back to a working version of the SVN with a snapshot. Happy to try to update and test if this helps

I’ve taken several screenshots that hopefully help.

just to clarify this is with Fedora 22 Server, PHP7 from Remi Repository with FOG version 7330 on an VMware ESXi server.
Previously it was working perfectly

Cheers.

[Sebastian Roth]

@RipAU You are seeing a lot of requests to srvpublic.crt which makes me think that this might be just a peek load while a lot of the clients redo the certificate stuff for secure communication to the FOG server.

Not sure if this was asked already but do you know which version you updated from that was perfectly working for you?

[RipAU]

@Sebastian-Roth

Yep just did a revert back to the previous version then. It is Fog Version: 6207
The CPU usage is a little higher than I’d like (fluctuates from 20% to 65% depending on the amount of traffic to the server, which is one of the reasons I was going to update) but otherwise has been working perfectly with no real issues.
Server load generally seems fine as well - load average: 0.88, 1.74, 1.55

[Wayne Workman]

@RipAU While the load is high, go into MySQL and issue this command:

SHOW FULL PROCESSLIST

and give us the output.

Maybe it’s time for you to tell us about the ESXi hosting machine, and the resources you have assigned to FOG.

Specifically, the host’s core speed and processor type. and how many cores you assigned to FOG, and how much RAM.

[RipAU]

@Wayne-Workman
Yep no worries.

The server is ESXi 6.
At this stage it has 1vCPU has been assigned with 2GB of RAM.
Server CPUS are Intel Xenon CPU E5-2630 @ 2.4ghz

Fog Server at this stage only has about 300 hosts (as soon as everything is smoothed out I was going to migrate the rest of our hosts over and this will be about 1100)

I can revert to the updated fog version of the snapshot after It was updated and run that command.
I’ll do that now.

Cheers,

[Wayne Workman]

@RipAU Oh… well there’s your problem. Fog sucks with one core.

give it at least two.

But don’t just bump it to two. Now you need to tear the whole thing down and start over - because your using a 32 bit operating system.

give it 2, and then re-install the OS.

[RipAU]

@Wayne-Workman

I can easily up the CPU and RAM - The server is the 64bit version of Fedora.

 4.2.7-200.fc22.x86_64 #1 SMP Thu Dec 10 03:28:47 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux

[Wayne Workman]

@RipAU said in After update of Fog Trunk - PHP Errors:

@Wayne-Workman
The server is the 64bit version of Fedora.

if the fog VM only has one core assigned, I promise you it’s not a 64 bit operating system. It installed as 32 bit. Why? Only one core. Impossible to execute 64 bit code with one 32 bit core.

-edit-
You’re basically executing 64 bit code on a 32 bit processor.

[Wayne Workman]

@RipAU Try to bump it to two cores. See what happens.

[RipAU]

@Wayne-Workman

Okay I’m confused.

So you are saying even though the ESXi host is 64bit CPU and I installed the Fedora x86_64 version of the OS
and the kernel is listed as x86_64 GNU/Linux.
Because I only allocated 1 virtual CPU It installed as 32bit?

[Wayne Workman]

@RipAU x86_64 compiled code can run on 32 bit or 64 bit systems. Because you have one single core assigned, you have one processor. One single 32 bit core. Meaning that one single core is running 64 bit os because of how that os is compiled, and the instruction set that allows a 32 bit core to run 64 bit code. It’s slower, but works. If you intend to take real advantage of a 64 bit OS, you need two cores.

Plus, besides all that, if your load is high, wouldn’t it make sense to have more cores? And why are you being stingy with cores? I have an old core 2 duo in my basement I use as a VM host - it runs 4 VMs, each with 2 cores assigned. It does fine for my purposes. But point being, cores can be shared.

[RipAU]

@Wayne-Workman

I’m happy to try a re-install and try to migrate the current database and images to an updated OS as I was going update to CentOS7 in future as I would like to be able to keep everything up to date.

I guess I generally will only give it a single CPU unless required as that was what I was told by a VMware engineer at a training session a few years ago and I can’t say I’ve even heard that only allocating a single 64bit capable CPU ends up running as a 32bit CPU on VMware ESXi.

[Wayne Workman]

@RipAU Pretty sure you can’t install CentOS 7 on a machine with one CPU. I’ve tried on an old Pentium 4 - didn’t work. I also tried once at work when I accidentally forgot to bump the cores from 1 to 4 for a VM - and CentOS 7 threw a nasty error when I tried to boot.

[RipAU]

@Wayne-Workman

No worries, I’ll look at spinning up an updated CentOS 7 machine to test the SVN with.

That said, I currently have a working version of CentOS 7 64bit with 1 Virtual CPU on the same ESXi and that didn’t have any problems being installed and everything is listed as 64bit by vmware etc.

I’ll give this a go over the next few days and report back.

Cheers,

[Wayne Workman]

@RipAU Well that is very interesting.